Those are new warnings that come with a more up-to-date llvm
Fixes#1066
Co-authored-by: eldritch horrors <pennae@lix.systems>
Change-Id: I305d0d810d12a5e8d31c7d89e0cdb3a82df71556
As far as I can tell, there's no harm to let know a nix-shell invocation
about NIX_LOG_FD being stderr.
Fixes#336.
Change-Id: Ifdb3591813251d4bc481158d28dc18e1489de72c
Signed-off-by: Raito Bezarius <raito@lix.systems>
As far as I can tell, there's no harm to let stdenv know that they can
write to file descriptor 2 which is stderr inside of a nix3-develop
shell.
Contributes towards #336.
Change-Id: I51dcbcbc19a1698a0d1255f9d943d9ebd15d115d
Signed-off-by: Raito Bezarius <raito@lix.systems>
Closes#1064
The culprit here is that `genGraphString` is only invoked with the
store-paths associated with the outputs of the derivation, so when
filling `dependents`, the `graph_data.find(p)` call would return the end
of the iterator when doing this for references to other store-paths.
As a result, the code wrote information behind the graph data-structure
causing a corruption. For me, this resulted in a SIGSEGV most of the
time and in a few cases in an uncaught `map::at`-exception as reported
by Niklas.
This patch changes two aspects of the original implementation:
* When filling `dependents` in the node-set, use `map.at()` instead of
`map.find()->second`. The latter doesn't make any sense and was the
cause of corrupting memory. The `at` would've made it far easier to
spot this in the first place.
* Filter out store-paths that don't belong to a different output of the
derivation when creating `outputGraph`. This variable is used on two
places, `genGraphString` and for topological sorting.
The latter already filters out store-paths from a different drv, so
this is happening now when creating the variable in the first place
such that `genGraphString` never ends up with corrupt data in the
first place. This is the actual bugfix.
Implemented a regression-test for this case to be sure.
Change-Id: Ie02144d89c32b0a776cb1ece0601d0229315ebc3
we'll need this to modify argv for socket-activated daemons. this is our
replacement for the old savedArgv mechanism that was unscoped and fucky.
Change-Id: Ie048eb8ea99f1c9cd627a051292c836c83197068
the parent daemon does not change any settings before starting a child,
so there's nothing we may want to change that is not already set by the
config file. this also doesn't prevent changes of the config file being
applied to daemons where we do not expect it since it'll only restore a
setting to the parents' value if the child also has an override for it.
Change-Id: Ic5a9ef13458c103ec9979cb187ba8d3ce5e1e719
Lix changed to C++23 in 4ea8c9d61¹, but the perl subproject wasn't
updated with it. This commit corrects that.
[1]: 4ea8c9d643
Change-Id: I5464138d206d87cf25762720b147487d6a6a6964
We offer the three usual options: verify, repair then delete and direct
people to report us corruption so we can fix if this is induced by us.
Fixes#447.
Change-Id: I0df61769d732d227333c206f312857c0593f7bce
Signed-off-by: Raito Bezarius <raito@lix.systems>
If the derivation does not start with D, do not return a simple
"expected string 'D'" error but a full error message.
This contributes towards #447.
Change-Id: Iee05f3918e4cc43e79f244ab2fd64a52cf2bb6d2
Signed-off-by: Raito Bezarius <raito@lix.systems>
Prior to I6a6a6964d2b5ad47ae5ea9eb11af9b6373ce2141 — `sudo nix
upgrade-nix` would perform direct store access.
This ensured a certain number of desireable properties for upgrading the
Lix binary itself.
We re-introduce direct store access for upgrading Lix binaries.
Fixes#1060.
Change-Id: I523c4d3023ed5fe9eff8fde9a266c56a0de47d8c
Signed-off-by: Raito Bezarius <raito@lix.systems>
Alternative to cl/4661 discussed in #1044.
It can be assumed that `$tmpdir/build-top` can be created safely without
any risk, this way, we don't need to reuse the random directory creation
primitive.
Fixes#1044.
Change-Id: Iec52477f3047fc40959b183c607312d5a40fc8c9
Signed-off-by: Raito Bezarius <raito@lix.systems>
Otherwise, loading Perl bindings fails early with[1]
undefined symbol: Perl_pad_sv at /nix/store/h2jsb5i4yfblr2f3ac2c7zpmlmj7zjym-perl-5.40.0/lib/perl5/5.40.0/XSLoader.pm line 94
Apparently, it's expected behavior by Perl that this symbol only exists
with `DEBUGGING` being set, hence it's used by the headers. However,
`pkgs.perl` from nixpkgs is apparently not built with `-DDEBUGGING`
causing this error.
Now, `NDEBUG` is manually unset after loading the Perl
headers rather than setting `DEBUGGING` causing the error mentioned
above.
I confirmed that this not only fixes the problem described above, but
running the Hydra tests with
diff --git a/perl/lib/Nix/Store.xs b/perl/lib/Nix/Store.xs
index dfdd64d28..14788266c 100644
--- a/perl/lib/Nix/Store.xs
+++ b/perl/lib/Nix/Store.xs
@@ -27,6 +27,7 @@ using namespace nix;
static AsyncIoRoot & aio()
{
+assert(false);
static thread_local AsyncIoRoot root;
return root;
}
still results in assertion errors.
Finally, added a small install-check that importing `Nix::Store` works
fine.
[1] https://git.lix.systems/lix-project/hydra/issues/69
Change-Id: I58521777eb0f94b766a9813aa4bbd06f9052bd35
As discussed in the F2 matrix channel, not being able to put the error
message inside of the thrown exception just results in one assigning an
additional variable to immediatly throw it. The original reason for
these rules to exist are that the message isn't printed twice, though
the line above the thrown exception is printed anyways in the
stacktrace, so we still get the error message twice but now with
additional work.
Hence this commit removes the said rules
Change-Id: I0b37c0b0861334703a5772c36353d31441f19dcd
So far, our ruff config was confined to the f2 package. This meant, that
when one added additional paths to the ruff formatter, those wouldn't
get the same rules applied as f2, resulting in inconsistent styling
thoughout the project.
Due to how configs are resolved, only the "closesed" pyproject toml is
considered. This means, we need to tell f2 to extend its configuration
with the base level one. Though no change is required for other parts of
the project, as long as they don't have their own pyproject.toml
Change-Id: I145c764e7b850194020b5560e1025f4aa80411ae
Nix daemon runs in a service context, usually with systemd.
If the Nix daemon unit adds `[Service] CacheDirectory=nix-daemon`, it should
successfully use `/var/cache/nix-daemon` automatically.
Instead, it uses `/root/.cache` right now, which is really bad.
By default, we add `[Service] CacheDirectory=nix` now which means that
caches are moved into `/var/cache/nix`.
Fixes#634.
Change-Id: I854b1045bfdce8f60110aea70bda1bf6657dfd7b
Signed-off-by: Raito Bezarius <raito@lix.systems>
Recently, in 54df89f601, support for mdbook 0.5
was introduced, including some logic to handle the `sections` -> `items`
rename. However, compatibility with 0.4's `sections` was only kept on the read
path, while writing 0.5's `items` unconditionally, which ends up in the bit
bucket on 0.4, effectively disabling substitution fully and leaving the include
directives in the final documentation. Restore writing into the `sections` when
they were there so that substitution works again.
Change-Id: Idd4d7653012660f3f7fc27f81f29b82d6a6a6964
Silly me forgot to add the dependency responsible for handling
`@pytest.mark.timeout` failsafes to kill a test if it takes too long,
which results in a warning when running f2.
This commit adds the required dependency and hence removes the warning
Change-Id: I281cad05bb0cf50208f72080cb2bfd8cbfb09d0c
Running the sandbox (as unprivileged user) requires all of user, mount and PID
namespacing. Previously, only the user namespacing support was checked.
Unfortunately, newer Ubuntu kernels impose restrictions on capabilities within
unprivileged user namespaces [1]; as observed by experimentation, in particular
(re-)mounting filesystems or changing mount propagation is prohibited even
inside a mount namespace, making it impossible to use the Lix sandbox. So any
tests requiring sandboxing should check for this breakage and be skipped if
detected.
`unshare --mount` performs the same problematic operations, so that's what gets
used for the test. PID namespaces are included as well for completeness, even
though currently no instance of breakage due to them specifically is known.
This fixes [2] for the functional test suite only.
[1] https://ubuntu.com/blog/ubuntu-23-10-restricted-unprivileged-user-namespaces
[2] https://git.lix.systems/lix-project/lix/issues/545
Change-Id: If95b527bf965ccb06371bbef7b6007666a6a6964
Builds using a diverted store strictly require sandboxing. Explicit disabling
by the user is already handled properly by force-enabling the sandbox anyway
(with a warning), but sandbox-fallback was still honored, predictably leading
to hilarious breakage (like the "confusing output" in [1]) when it actually
triggered. Cause a hard failure instead.
[1] https://git.lix.systems/lix-project/lix/issues/545
Change-Id: I61ea32c3bd8bca4e5ba278312b0a49326a6a6964
Builds using a diverted store strictly require sandboxing. It therefore makes
sense to automatically enable it even if otherwise explicitly disabled by the
user. However, they should still be informed that their choice is not being
respected.
Change-Id: Ia2cbdc24a4c55d8ecc49094e35e698e16a6a6964
Disabling build sandboxing is a significant change to build isolation. If this
is done automatically due to namespacing issues (and sandbox-fallback not being
disabled), show the user a loud warning without the need to opt into a lot of
other log spam.
Change-Id: I52daccdd6934873234223c798ca3dcff6a6a6964
This contains the depth of nix shells nesting.
Fixes#826
Co-authored-by: Qyriad <qyriad@qyriad.me>
Change-Id: If584c9d02730d6c857636dafdeab0c01f4ec8e0f
If the NIX_PATH shadows internal files, this will often break things,
hence we forbid it.
Fixes#998
Change-Id: I70e5d389532ada1c9f910c60281abe565e3ce6bb
Some of the tests depended on the evaluation order chosen internally by
nix-eval-jobs, which will break when that order changes. In practice the order
is not guaranteed, and already not deterministic with multiple workers.
Change-Id: I2e85d096f4439a8075a6f4a08d31c0f16a6a6964
GCC still ICEs on coroutines, while Clang works properly. Use the latter, just
like for Lix itself.
Change-Id: I802d6df7279550a5f5355e6c1ea0533c6a6a6964
Previously the collector-side implementation of the worker interaction protocol
was a bunch of spaghetti code. Split it apart at places where it makes sense to
be easier to follow.
Change-Id: If3cc6b6fbf289dd778856b40b55316c76a6a6964
As the worker processes are now properly spawned using fork-exec, they no
longer inherit the stack size from the non-main coordinator thread, but get a
main thread with reasonable stack size on their own. For this reason the
coordinator threads can use the default stack size, and turn std::thread can be
used as the only reason for the custom wrapper was to enlarge the stack.
Change-Id: I1192474885abe9b0625ac483840b6e1a6a6a6964
Similar to how 7b37d5ea6a switched the subdaemons
from fork-only to fork-exec, do the same for the nix-eval-jobs workers. This is
the promised change that causes the signal handler thread to appear.
Change-Id: I590a5eb6bddfb546a331a01191fd788b6a6a6964
The retry added in 4c5efd4548 has never worked
properly, as releasing the pid the first time causes it to be -1 and thus
waiting for a random child on retry. While there is a good chance that this
really catches the crashed worker, this is by no means guaranteed, and it could
just as well have caught different restarting worker. Wait synchronously
instead. The worst possible outcome is a hang instead of an error message if
the worker manages to close its pipe but then fail to exit, which is not
supposed to happen in the first place.
Change-Id: I4e4c3f532ec15c3118f6f77821ab820a6a6a6964
warning: embedding a directive within macro arguments has undefined behavior
????? CLANG? why does this take -pedantic to show??
Change-Id: I6f4a9f3db1b89448f9f5d00d82b551e46a6a6964
This change brought to you by my uninstalling go but forgetting to
remove ~/.local/opt/go/bin from my PATH.
Change-Id: I937675bab11ac7e0678c896cba78204b6a6a6964
Maintains compatibility with mdbook 0.4.x.
Includes comments for what to remove once 0.5.x is the only we care
about. Some other changes technically could be changed at that point,
but currently serve to enable universal support of 0.4.x and 0.5.x
Fixes#1051.
Change-Id: Ic5b405038d180bcd357bbd9e5716879e0c26e5f5
In the same vein as be18b7dc2¹.
Before:
error: syntax error, expecting ';'
After:
error: syntax error, expecting ';' to end binding
[1]: be18b7dc25
Change-Id: Iffbcb113d2b892a50c646c9875e970376a6a6964
Lix blocks SIGPIPE, and std::cout ignores EPIPE (actually it probably stashes
an error code in some flag that no one ever reads, with the same end result).
Consequently, nix-eval-jobs would waste resources by continuing to evaluate
even when the reader interested in the results has long gone away. Instead, use
writeToStdout, which throws an error on EPIPE, leading to the process
terminating as desired.
Reported-by: Winter <winter@winter.cafe>
Change-Id: I962c09bab582a8ed27dd41c01b1519876a6a6964
Right now, the worker will not receive interrupts, as it has no signal handler
thread due to the way it forks, but this will change soon. Prepare to handle
worker interruption properly, by letting the worker exit successfully in this
case and adjusting the coordinator to handle this exit gracefully.
Change-Id: I9f79670b0d2004e7e2d8bf36cf67108e6a6a6964
For some reason that eludes me, it used to only print the last attribute
instead of the full chain.
Change-Id: I2f5f6e85e2b1ab3ab20a9c99bf06d1e45a4d09bb
`e`, `ae`, `j`, `jAttrs` … holy fuck this function was so much worse
than it needed to be. This refactoring includes:
- Descriptive variable names where possible
- More code comments indicating what even is happening
- Shuffled the control flow around to be more linear (move early return
conditions up, etc.) and have less rightwards-drift, to increase
readability
- Extracted `mergeAttrs` function for future changes and readability
Change-Id: I7253b47ce4910fdc67aa7d6edff2f3c0bb8fa55b
In one of the cases it pointed to the first instead of the second
occurence, which is a bit confusing
Change-Id: Ie508d1f84feb434708804d45aa8a9c8b1e3c5f69
How long do we have the value printer already? It's time to stop
concatenating strings like it's 2005
Change-Id: I3f5074de2439a1ad78af94de877bb141bc9f1d82
When using `builtins.derivation`, the `outputs` list can be omitted, and
in that case, it defaults to `[ "out" ]`, set this default when building
the shell derivation for `nix develop`
Fixes: #556
Change-Id: I4dc77d060f30ee499e186ec919de30bd3e5c53bb
Commit 5dc847b47b introduced usage of
std::chrono::milliseconds, which prints with the unit, leading to its
duplication. Remove this duplication again. Specifically, the explicitly
printed one is kept, due to better typography.
Change-Id: I724ea6be6f2a1349fe799602c3429f096a6a6964
We have done it!
The functional/lang framework has fully been migrated to functional2 :D
closes: #856
Change-Id: I63ad8d7dbcd9b5267ca04af68df73b1ffa3d6461
After multiple cycles of deprecation, we now use the lix repositorty as
a default, when no prefix is defined
Change-Id: I29b445fee39fb20c0ff024e3b95402f459144b62
Added builtins.warn` which takes two arguments: a message that is
displayed as a warning during evaluation which must be a string and a
value that is returned from the expression.
The next commits add new settings to control the behavior of the new
builtin: `debugger-on-warn` allows the user to start the debugger and
`abort-on-warn` aborts evaluation with an error.
Unlike upstream, I chose not to mark evaluation warnings from
`builtins.warn` as distinct from other warnings because that breaks the
commonly expected logging format `level: message`.
Co-authored-by: Qyriad <qyriad@qyriad.me>
Upstream-PR: https://github.com/NixOS/nix/pull/10592
Fixes: https://git.lix.systems/lix-project/lix/issues/579
Change-Id: I8658c88e5c27952b65e8b9f5525a572e0680cc1f
There are currently 4 different ways to run garbage collection using
lix:
- `nix-collect-garbage`
- `nix-store --gc`
- `nix store gc`
- (using the daemon directly)
As they were written all at different times, their output varies (and is
broken in some case). This unifies the display of informations in the
following ways:
- The list of paths in the results is always printed (in the nix3 cli it
is hidden unless `-v` is passed)
- The number of paths in the result set is always displayed
- The size of deleted paths is only shown when actually deleting things
(as it would have been 0B in any case)
Fixes#905
Change-Id: I40d9ec7c6d76795f6c6dd30df196d1e855bdb9db
This also allows removing a static initialization of a primop.
Co-authored-by: Eelco Dolstra <edolstra@gmail.com>
Change-Id: I74da68205630f658b239e4327bf5e27bfc1b91da
This is done at the start of the command call, which allows not building
expensive things if the arguments are incorrect anyway.
Fixes#687
Change-Id: Ic924fe2115cf760684c6fdf7987279e96ab00286
the newly added f1/lang tests is required for the f1/lang framework
dismanteling to happen in a separate commit
Change-Id: Ic419c515262294c51a46d1513daa7848e4b71405
While the err file should always be empty, we prefer not have (easily
avoidable) warnings in the log, in order for actual warnings to be
spotted more easily.
Additionally this way no additional changes are required in case they
make use of some depreacted features in the future.
Change-Id: Ie078e2a851b2035d839f6b95d4188e475eb96b2d
autoargs get their own folder and are not allowed with the other
function args tests, just like they deserve their own ring of hell
It shall be noted though that this test is particularly pointless, as it
merely tests the normal argument passing (which is fine), and not the
✨magic✨ autoargs actually provide
Change-Id: Ia2dc00754a13c02e4926200141f90bc797820a15
That one is a bit confusing, as I renamed the old `in` test to `in-2`
(file identities in Git when)
Change-Id: Id0e17de414dbc1b1b616d456175704f98657a16a
Call it a bug, call it a feature, but the old testing framework
ungracefully shits itself when it has no tests to run … in other words,
we finally migrated all the parse-fail tests 🎉
Change-Id: Ibd7f1c04d9a396a20f14361af7924f0074d1ac23
The second test has also been fixed. I'd do it in two commits if
functional1 wasn't such a PITA (guess why we're doing the migration …).
Basically, the introduced syntax got changed later on, but the test
never got updated, and back then tests didn't test the output so nobody
ever noticed it now failing for a wrong reason.
Change-Id: I5b66127d507c65676b8dd4a82d1e1f2857fef145
This really looks like one of the first tests ever written, it doesn't
seem to be actively testing anything interesting
Change-Id: I442840db932203c25da2d8400ada791f8aed04ce
This merge commit returns to the previous state prior to the release but leaves the tag in the branch history.
Release created with releng/create_release.xsh
Change-Id: Ida5959473ec469dd5e5e977ebdc8aab6e694b0de
https://github.com/NixOS/nixpkgs/pull/451579 PR
enabled c-aresSupport for curl on darwin which ended up breaking
DNS resolution in FOD sandboxes:
```
nix-run> exporting https://tangled.org/@weethet.bsky.social/nix-run (rev 73d7bf6b58848fb8f42e3a69816e0847f041c689) into /nix/store/m4m951648wmipxgwrgsml9gzjwfpfhm7-nix-run-73d7bf6
nix-run> Initialized empty Git repository in /nix/store/m4m951648wmipxgwrgsml9gzjwfpfhm7-nix-run-73d7bf6/.git/
nix-run> fatal: unable to access 'https://tangled.org/@weethet.bsky.social/nix-run/': Could not resolve host: tangled.org (Could not contact DNS servers)
nix-run> fatal: unable to access 'https://tangled.org/@weethet.bsky.social/nix-run/': Could not resolve host: tangled.org (Could not contact DNS servers)
nix-run> fatal: unable to access 'https://tangled.org/@weethet.bsky.social/nix-run/': Could not resolve host: tangled.org (Could not contact DNS servers)
nix-run> Unable to checkout 73d7bf6b58848fb8f42e3a69816e0847f041c689 from https://tangled.org/@weethet.bsky.social/nix-run.
```
with these sandbox failures:
```
deny mach-lookup com.apple.SystemConfiguration.DNSConfiguration
deny file-read-metadata /private/etc/hosts
deny file-read-data /private/etc/hosts
```
We allow those so that DNS resolution work fine
Change-Id: I9102293691972feb085adf8e9b1ad915bb3a36ab
This adds the enablement code to support HTTP/3 if the user requests it.
We leave it disabled because h3 is not onpar with h2 performance.
Change-Id: I1fd3d4c97b972dcf36bccacc6c9a8290e22b31e0
Signed-off-by: Raito Bezarius <raito@lix.systems>
Channels have moved from `nixos.org/channels` to `cnannels.nixos.org`.
This udates all relevant links (excluding release notes) to use the new
canonical URLs and replaces HTTP with HTTPS.
Fixes#1031.
Change-Id: I212821c44ac5e482c8e9eaa415c7d8ee17ff8341
Signed-off-by: adam <me@adamperkowski.dev>
URL literals were deprecated in 278fddc317,
effectively stabilizing the experimental feature. Enabling it has no effect any
more. Remove it.
Change-Id: I8fbca03b1a2be6a8cddfce644043ac0a6a6a6964
We start this section with shortcomings of unsandboxed builds.
Fixes#1018.
Co-authored-by: eldritch horrors <pennae@lix.systems>
Change-Id: Ieb17e4340beab0c1197951813ae602de453a3fd9
Signed-off-by: Raito Bezarius <raito@lix.systems>
This is unnecessary because Darwin builds already cannot create
`set{u,g}id` files due to the minimal sandbox policy, because we can
forbid messing with the top‐level build directory directly in the
sandbox rules, and because Darwin builds can trivially avoid isolation
through temporary directories right now anyway.
This does regress the build directory isolation for builds with
`sandbox = false`, but I can’t imagine that mattering given the
above. The sandbox change prepares us for a world where we close
off shared temporary directories for `sandbox = true` builds and try
harder to achieve proper isolation on Darwin, but probably doesn’t
have a meaningful security impact one way or another for now.
With this change, we get down to 41 byte build directory paths on
Darwin, a ~2–3½ byte improvement over the old status quo. We can
also restore the 0710 permissions on Linux.
Change-Id: I6a6a6964a681c0365241fe7234831db656b76799
The minimum build directory length on Darwin with default settings
when we were still using `/tmp` was 22 bytes. Deriving build directory
names from my local store, the median and mode were 43, the mean was
around 44½, and the maximum was 127.
The switch to `/nix/var/nix/builds` over `/tmp` added a 15 byte
penalty, and the additional `/b` directory added another 2.
Now that we use opaque build directory names, the length is 48, so
we’re still at a ~3½–5 byte penalty over the previous status
quo. This change brings us down to 43, matching the previous median.
Note that these calculations do not take into account the fact that
`/tmp` is a symbolic link to `/private/tmp` on Darwin. Anything
that was canonicalizing paths would have had an additional 8 byte
penalty in the previous status quo that is not applicable here,
so we may already be ahead even without this change. If the more
opaque directory name here is undesirable, then that factor could
potentially help us squeeze by without. Alternatively, in combination
with dropping the `/b` on Darwin we could use `/nix/var/nix/bld` or
similar, but I feel that the paths in general are sufficiently opaque
that it should be okay to go with the shorter option here. Given that
some projects already had to reduce filename lengths to avoid this
limit even before the recent changes, I think it is best to try and
improve on the previous status quo.
Note that `/nix/var/nix/builds` will unfortunately not be cleaned up
on Darwin. However, we don’t clean up the directories inside it on
Darwin anyway, so hopefully that’s okay for now?
Closes: #913
Change-Id: I6a6a6964bffce7194bcddcaefb4c4a37569c7df5
We always use the default temporary directory, because
`createUniqueDir` has an interface nice enough to use directly for
the few bespoke uses.
Change-Id: I6a6a696450b7c0a0bd76655632fb14d7c5e38199
This makes the paths more nondeterministic, but more reliably unique,
and lets us remove the retry loop.
Note that this adds random entropy to the build directory visible
inside derivations on Darwin and unsandboxed Linux. It was already
non‐deterministic in the presence of concurrent builds and similar,
but now we can reliably expect it to be different every time. On the
whole I think that’s a good thing, as it is impossible to ensure
a single consistent build directory and derivation outputs should
not depend on it.
Package reproducibility isn’t great on Darwin to begin with,
though, and the reproducibility bugs this will turn up in packages
will be more urgent to fix than when the build directory was mostly
consistent. A quick survey of my local store shows that many C, C++,
and Rust binaries contain build directory references, likely due to
use of `__FILE__` and its equivalents; non‐binary offenders include:
* Install logs included in the Rust and Cargo bootstrap compilers
* Example errors in the Rust documentation referencing build paths
* Configuration information installed with CPython itself
* Python 2 metadata from resholve’s closure
* Cython metadata
* Generated headers in Facebook libraries referencing source paths
* Generated CMake files in Facebook libraries referencing source paths
I haven’t built that much in this store since the last GC, so this is
probably only a small sample of the problems across the tree. These are
all instances of <https://reproducible-builds.org/docs/build-path/>,
though, and should probably just be treated as general reproducibility
bugs outside of contexts like the Linux sandbox where we can normalize
them away entirely.
I have implemented away build directory paths for C/C++, applied some
additional fixes for non‐`__FILE__`‐related issues in binaries
from ATF and LLVM, and fixed the derivation bug causing the CPython
3 issue, and will work on upstreaming these changes. Rust is working
on the problem upstream, with some temporary workarounds we can
potentially apply in Nixpkgs for now. The rest will require some
distributed effort.
Change-Id: I6a6a69645b4915c56c0fdef904aa81684e4136c6
This does change the behaviour when the global temporary directory
does not exist, but other uses of the global temporary directory are
already broken in that circumstance, and it should be fixed centrally
if the use case is considered desirable. The logic was not present
before the recent churn around build directories – it was added now
that Lix is taking ownership of the build directory in the store –
so this should not be a meaningful regression.
Change-Id: I6a6a6964e345ea6803226c5ad759e836de7cb0ed
There is now no risk of race conditions on a system with a functioning
entropy source, and the bespoke prefixes are either redundant to the
default or unnecessary.
Change-Id: I6a6a69641211c6bb979ea48ad30aecb1a53d03f0
Relax the constraints on keeping the exact same filename format to
provide a more robust source of entropy with a simpler interface
(as previously suggested by eldritch horrors). Using 128 bits of
OS‐provided entropy ensures global uniqueness and allows us to
skip any thought of gracefully handling the case where these files
already exist.
My microbenchmark that repeatedly constructed paths like this and
printed them out showed that this takes about 1.23× the time of
the previous implementation, both taking on the order of a couple
microseconds for one iteration. Since everything that uses it is doing
things more expensive than printing to standard output, the actual
performance delta is likely to be lost in the noise. If it somehow
becomes a bottleneck, it can be optimized without sacrificing the
guarantees by reading from the system RNG only to seed a thread‐local
CSPRNG like [ChaCha8Rand], but I think that’s very unlikely.
We also tweak the recommended way of creating a temporary file inside
a directory in anticipation of later changes, and rename the `suffix`
parameter to `prefix` (it’s a prefix to the random characters and
a suffix to the root, but this way is more consistent).
[ChaCha8Rand]: https://c2sp.org/chacha8rand
Change-Id: I6a6a69648502c746d13d8c3bd2768cbbf1b90466
base32Encode now takes std::span<std::byte>, with a base32EncodeStr
convenience wrapper which takes std::string_view.
Co-authored-by: Qyriad <qyriad@qyriad.me>
Change-Id: I6a6a6964f799dc84ecbfb55c7ca03a064cff71d9
the change to shareable thunks also removed a few cases of pointer
equality checks that allowed structures containing functions to be
considered equal to other sets containing the same functions, even
if the sets themselves were pointer-equal themselves. *so* busted.
Change-Id: If87fdab658f9037ce2a654f69a9e3da6ae2f53e5
we can't change the protocol to allow daemons to do this, and we should
not try to guess what the `auto` store uri means depending on whether a
command was run by root or not due to copious side effects and not even
being able to tell whether the `auto` store uri was given explicitly or
not. while `auto` may *technically* allow this via its naming we should
resist the urge to add a hack and fix the underlying protocols instead,
especially since repairPath should be a rare, superuser only operation.
fixes#888
Change-Id: I1b53245db226199f827a89a237a2ab9907c3f766
if curl fails during a transfer and said transfer is piped through
libarchive we see very strange errors (#1009). in such cases we'll
want to prioritize reporting the source errors to allow debugging.
Change-Id: I8cc231257eca5b749471ecd38df15071170decc2
The test would fail if the identity cannot be determined
(as is the case e.g. when using auto-allocate-uids under darwin).
Instead we use the uid when the user name is not available.
Fixes: #1025
Signed-off-by: hexchen <hexchen@lilwit.ch>
Change-Id: I12f35840b78f989b5fd09bbd44c833dc6a6a6964
This helps users to debug whether their current profile symlinks are
correctly set.
Expected outputs look like this:
```
❯ sudo ./outputs/out/bin/nix doctor
[snip]
[PASS] All profiles are gcroots.
[PASS] Client protocol matches store protocol.
[INFO] You are trusted by store uri: local
[FAIL] Error: current generation cannot be discovered for profile:
'/nix/var/nix/profiles/default'
```
```
❯ ./outputs/out/bin/nix doctor
[snip]
[PASS] All profiles are gcroots.
[PASS] Client protocol matches store protocol.
[INFO] You are trusted by store uri: daemon
[PASS] You have 28 generations for profile
'/nix/var/nix/profiles/per-user/raito/profile'
The current generation number is '290'
```
Change-Id: I50c69cbeac3291d668f4c2332803411579adc944
Signed-off-by: Raito Bezarius <raito@lix.systems>
Co-authored-by: Qyriad <qyriad@qyriad.me>
Inaccessible-but-extant PATH elements can't be executed anyway, so we
can skip them as if they didn't exist, instead of erroring like before.
Change-Id: I9288c3ecb6768171e62a3088122d98421558eb03
In d2696cdd¹, libstore was made to prioritize NIX_BIN_DIR/nix (over
/proc/self/exe) if it exists. But `pathExists()` throws an exception if
lstat() returns any error other than non-existence.
Because this check happens in libstore's Settings constructor for the
static-initialized global `nix::Settings`, this case would abort Lix
well before main().
[1]: d2696cdd1e
Change-Id: I62e44c8a33b28c234900df3e4790c1badefee358
So far it was impossible to put absolute paths into files, as one
wouldn't know whre the test would be placed.
This commit adds a new Fileish variant called `EnvTemplate` which uses
the given string as a template and replaces the `@ENV_VARIABLE_NAME@`
placeholders with the according value of the environment variable.
This way one can use `@HOME@` or `@TEST_ROOT@` to build absolute paths
Change-Id: I425cb6408dceb8d7f26d136ace4ac98b1ca31ec3
Since we have async, we are unleashed and therefore we started to cause
people to run into system limits like number of open files.
We introduce an async semaphore which uses max(25 % of max open files,
max number of cores) which should lead to 256 maximum concurrent copies.
We believe that the copy operation is the only one that can cause issues
given that the daemon already runs with obscene fd limits.
Fixes#1022.
Change-Id: Iec433d10d5c5003962ea749cf8e32bafb314f0d9
Signed-off-by: Raito Bezarius <raito@lix.systems>
In case of empty messages, it is good to print the raw error code.
Additionally, we print request IDs which can help users to reconcile
what happened with the service provider.
Change-Id: I4d83c011c1b7a5514e3d1b21123df38308279044
Signed-off-by: Raito Bezarius <raito@lix.systems>
Completion status can be updated in two ways:
- progress callback
- transfer status update callback (esp. important in multipart
transfers)
It's guaranteed that one of the two will provide progress for our
promise.
This fixes#945.
Change-Id: Iac9e92df34f82fbc1facae7a411ccabf1688da52
Signed-off-by: Raito Bezarius <raito@lix.systems>
This fixes problems with S3 implementations that have mandatory payload
signing, e.g. Garage v2, AWS S3 itself, etc.
The problem manifested itself when the AWS SDK threw some error codes 99
(NETWORK_CONNECTION) with no messages and retried until it finally came
to a moment where the state machinery decided to send a
`Transfer-Encoding` header in a `Content-Encoding`/`Content-Length`-set
request with signed headers (even though payload signing is disabled),
causing the server to reject the transfer and crash the copy.
I did not debug super far what went wrong in AWS SDK, but I can confirm
this change makes transfers possible to finish with Garage v2.
Change-Id: Icc9e6a9f2afb0d760cf2d1e27816decd385a1d85
Signed-off-by: Raito Bezarius <raito@lix.systems>
Calling `nix.nix` with an empty argument list, would crash on darwin due
to a Index Out Of Range Error, as the build="auto" functionallity would
try to access the first element of the provided arguments (second
element at the place of call) resulting in a crash
Change-Id: Ia678109808b3f75a30182114baa421cedaa8759b
libarchive's xz offers single threaded xz compression which is very slow
and provides ~10-20Mbps compression speed in addition to maxing a core.
In exchange, it achieves optimal compression ratios among all our
compression methods.
Nonetheless, xz prevent the saturation of 1Gbps+ connections and slow
down significantly decompression for end users. As these connections and
faster hardware is becoming prevalent for cache servers and clients, we
offer to default to zstd.
Lix is a "compress once, decompress many times" application. To avoid
incurring a high penalty to end users very sensitive to compress ratio
(very slow Internet connections), we dampen the consequences of
switching to zstd by increasing the default zstd level to 12.
On one example, xz will compress a 4.4GB file to 632MB, zstd on 12 will
compress it to 775MB, that is a ~18 % increase over the optimal xz
compression. zstd took 18 seconds to produce this file.
Increasing to level 14 leads to a 773MB file while taking 37s.
Increasing to level 16 leads to 735MB file while taking 66s.
Finally, xz took 77s, so a 50 % reduction in time taken to compress in
exchange of an increase of 18 % of the compressed size.
This change will reduce issues encountered in #945 but is probably not
the root cause.
References:
- https://discourse.nixos.org/t/switch-cache-nixos-org-to-zstd-to-fix-slow-nixos-updates-nix-downloads/23961
Change-Id: I7beda2bf2c1fed146dcb797b8f85dc290c486ab2
Signed-off-by: Raito Bezarius <raito@lix.systems>
if a promise for an async s3 operation is cancelled is inner fulfiller
will be destroyed, but the s3 operation may still be running. once the
operation finished and attempted to fulfill the promise we'd segfault.
Change-Id: I603c75d6cc39a1aee465b3ceee06b638f7f7f596
As we saw actual users running into them throughout the debugging of
#920 and #1014.
It's best to document and point end users to this so they can take
measures for themselves.
Change-Id: I7d08407f4354055bf65fc6dd7d1624c5a9304402
Signed-off-by: Raito Bezarius <raito@lix.systems>
Fixes#1014.
Address family detection is performed by Pasta, we influence Pasta into
enabling dual stack by listing both an IPv4 & IPv6.
Change-Id: I4121e0eae7d5185c287957106adc36edf3f82a40
Signed-off-by: Raito Bezarius <raito@lix.systems>
Instead of letting inner shell perform the cleanup, we will fork/execvp
and finish the work ourselves. Once an interruption arrives, we wait
again and return the exit status code of the inner shell.
This idea was suggested by eldritch horrors.
Fixes#1020.
Co-authored-by: eldritch horrors <pennae@lix.systems>
Change-Id: Id38af7ac33874aa1c4b7ca9c6123e3cbd8c3cd8c
Signed-off-by: Raito Bezarius <raito@lix.systems>
This was forgotten which should have been part of 64b9247f.
Change-Id: I3cf83d8699f2c1963ce530bbea6a511075202a79
Signed-off-by: Raito Bezarius <raito@lix.systems>
Legacy commands "main function" will now return an exit code.
Fun fact: `main_nix_build` was the only one not returning integers
already.
Change-Id: Ia43a16c3c3fb9a670e8889aefc4ee9b6528a7df4
Signed-off-by: Raito Bezarius <raito@lix.systems>
This also prevent `legacy` to return nothing as it's void.
Change-Id: Ic797544a59b04b41ad9e1c46af4cad3a1cb6fdc8
Signed-off-by: Raito Bezarius <raito@lix.systems>
realistically only runProgram is ever called for interactive reasons,
and even those calls seem to be rather ill-advised in many cases. the
chance of multiple interactive processes interfering with each other,
whether for input or for output, must be very low to make this in any
way reasonable: if e.g. git calls ssh for multiple fetched inputs and
ssh requests passphrases for both we can otherwise not guarantee that
*any* input is routed correctly. misrouted output is merely annoying.
Change-Id: I794e3fdf0a3238cb9292003a89ac267f0de7a939
Eval errors will now print a simple, no frills chain of involved
derivations at the bottom. For example, trying to evaluate
`pkgs.xonsh.override { python3 = pkgs.python2; }` has the usual
Package ‘python-2.7.18.8’ in /nix/store/9v6qa656sq3xc58vkxslqy646p0ajj61-source/pkgs/development/interpreters/python/cpython/2.7/default.nix:398 is marked as insecure, refusing to evaluate.
message, but now also includes the following:
note: trace involved the following derivations:
derivation 'xonsh-0.19.9'
derivation 'python2.7-xonsh-0.19.9'
derivation 'python2.7-setuptools-44.0.0'
To give the user information about why the erroring derivation was
involved in the first place.
We would like more structured information in the future, but this should
still be a significant improvement.
Change-Id: Icf6da52abd0a043cfb63943bf0b0c160c21ee59e
the wrapper is needed by transfer streams to restart a failed transfer
if desired. curlFileTransfer itself is more of a fancy handler for the
thread we're dedicating to curl io handling. the thread will stay with
the multi handle for now because quit handling needs to stay there. we
could have CurlMulti keep only a flag, but that does not help us much.
Change-Id: I99550f0bbb635b75898ca7260f08275df86050e3
we have anonymous namespaces and no other scoping requirements, we can
move it out. this will also make it easier to move *other* things out,
such as the transport streams and all the shared multi state handling.
Change-Id: Iafa1eaeb69e096e98c4667f0531b17d5fe7c78a6
waiting for them to finish leaves us prone to deadlocking, e.g. if some
active transfers is blocked and hasn't been destroyed before our thread
was asked to exit. aborting transfers is the safer option here; even if
we cause use after frees we'll at least have a crash, not hang forever.
Change-Id: I333dcced810e8a85e2ded1f377962a0176798098
If BUILD_TEST_SHELL is not set or empty, None or the empty string respectively
should not make it into the PATH. Ensure this property.
Change-Id: I4ce9b0c06c407b465308b63b9cb64e7d6a6a6964
this is still somewhat experimental and should be considered in flux. we
will have to nail down a logger interface once we start moving the store
protocol to rpc, but until we do that we can use build hooks to test it.
Change-Id: Id20cd346c9520f45871799c31b0af040adde56ef
extracting the size-specific log handling into a new sink deduplicates
the size-limiting code and lets us clean up the derivation goal a bit:
a number of log-related fields we kept only because previous ownership
semantics were a lot more ad-hoc can be moved into the new sink class.
Change-Id: Ie770a276597c84f9d928a9624c0f0de86adfe815
we'll want to remove handleRawChildStream from DerivationGoal, but the
timeout handling is still important. the wrapper is thus useful, but a
more generic way to pass it a promise to do the handling is necessary.
Change-Id: I55bf041d51e4f8a8b8291deae053423d06724329
we never need the connection or the rpc client directly after the hook
is constructed, only the hook instance rpc object. wrapping them using
kj primitives makes this more explicit and prevents destruction errors
Change-Id: I7e0b14d7e365c2e001273e04c0e741ba4afb12f2
this makes it possible to wait for the hook to exit in one place,
process its exit status for some purpose, and later process it in
another place for a different purpose. rpc will need this to shut
down a hook cleanly after the hook has completed its assignments.
Change-Id: I5af2d5aac4b02a0a200d720d0e6f0b5df5496aaf
only LocalDerivationGoal needs these, and once the child is done
running (or is begin torn down) we can just destroy all of them.
Change-Id: Id4d9f0105aaf33d6736c435bb891d3f9af8e7894
wrapping the wait in a timeout at the call site is insufficient because
a timeout would not relock the mutex and leave the raii object invalid.
Change-Id: I0be2aaf6b3c9966070fb0f1deaab048765fd7af3
it seems like pytest is updating the message printed when skipping a set
due to no parametrization
which leads to breakage on staging-next.
this commit changes the expected message to allow both old and new skip
message
Change-Id: I34d11ad84d7075911d8229fbc89889bb880a27c9
This partially reverts commit f68233ec43.
while more modern capnp in *does* have the build system changes, it
does *not* have the patches. and those patches are rather important
to us; the clock patch because our CI machiens don't behave the way
capnp expect and the nodiscard patch because it *is* a bugsquasher.
Change-Id: Ic273777f09fd8e5e7001f815db8998dae2dd7b88
The prospective callers of this should probably be doing something
smarter or more abstracted to begin with, but this is useful as an
incremental improvement for call sites with existing `makeTempPath`
logic in the face of filename length limits.
Change-Id: I6a6a69644292f5bbf984a1df90192e06c6022b53
LocalStore::optimisePath() was made asynchronous in 38b75b736¹, but we
lost warnings for not awaiting kj::Promise in f68233ec4², since our
capnproto no longer has the [[nodiscard]] patch.
[1]: 38b75b7367
[2]: f68233ec43
Change-Id: I2363396949f40dd43eae08cc76be9ede6b488cf7
this is important for rpc error reporting: if the peer disconnects we
want to throw a useful error that can e.g. let pending builds finish,
not kill the entire process because kj errors are non-lix exceptions.
Change-Id: I41ebe10ad7fd65d886238ce52eeca1fa243b1bb2
this removes another file descriptor from the build hook. we are now
using only rpc to communicate between build hook and daemon, setting
the stage for a future world in which we don't even need build hooks
Change-Id: I706d9d3f2a420abd42fc7762c205931a45d3c8c5
interpolation of the remote store name into activity text is now done in
the build hook, so we no longer need to send it back to the daemon. this
also reduces the dependency of the daemon on how *exactly* remote builds
are done a little bit, and maybe in time we can make these fully opaque.
Change-Id: Ie51c4f2cd107350fec4ad0a8ef83621ed86db6a2
only local builds can now have a builderOutFD, remote builds only log
via json streams. these two kinds of logs have different requirements
and each is only needed by its respective build kind. splitting these
apart thus makes sense, though ideally we'd also split DerivationGoal
into a RemoteDerivationGoal to clean up the rest, but that will wait.
Change-Id: Ib577537266d1160355ab9c44b4604ebda87a7d04
treat it like ssh-ng instead, and have the build hook do the translation
of ssh stdout to the json log message steam the derivaiton goal can use.
since the ssh-ng path also handles all store urls that aren't legacy ssh
we now have a single logging system that handles every remote build kind
equally, without requiring fd passing into the build hook. this is later
required to rpc-ify the log stream emitted by build hooks to the daemon.
Change-Id: Ifb522eb8a9745029050f16b1b3b3601a6ddac748
When the maximum memory size is chosen too large (relative to the free memory
usage), so that the evaluation workers don't fit into the free memory, they
will start being swapped out. Further increase in memory usage will not be
reflected in the RSS, so they will not exit, and instead fill up all swap too
until the system runs out of memory.
The BDW-GC keeps track of its heap size, including any parts of it that may be
swapped out. For this reason it does not suffer from a similar problem. While
not all memory usage by the evaluation worker is accounted for by the GC, a
simple affine model for the overhead works reasonably well in practice. The
specific parameter values were determined using nixpkgs evaluation, and appear
to work quite well also for somewhat different workloads (like the `hydraJobs`
of Lix, which include a bunch of NixOS system configurations). So this is what
gets used.
When Lix is configured without the BDW-GC, of course this approach cannot be
used. In this case the old strategy is retained.
Change-Id: I6cb5f0a9b4ceda9dd14be165dda108cd6a6a6964
Includes link to the Buildkite SSO login, which is currently not visible
anywhere else in documentation.
Change-Id: Ic3ec6412127b2ac0df7e6efc4ec6020cc9d8b34f
These are a footgun and are not acceptable in functional2 due to thread
safety, effects on other tests, etc.
Co-authored-by: Commentator2.0 <lix@crystal-cavern.systems>
Change-Id: I8d7285061eaa9bab27edd52f3646024c8cf605e5
When providing both a `.err.exp` and a `.out.exp` for a lang test
**not** containing a toml, the test was collected twice and hence threw
a duplicate id error.
This commit adresses this issue
Change-Id: Ia781d69e0f1db2809a30192d6cd0a98861e6cc32
this restriction was added in 4af2611bd1
and extended later. it makes little sense to allow *one* activity type
specifically and drop all others on the floor, especially since broken
builders can create transfer activities with ids that would be used by
other, "real" activities. the only thing achieved here is to drop logs
written in json format in the build sandbox, which does not help much.
we may revisit this restriction later during other log-related rework.
Change-Id: I8bda494083877b71a2f958470fa52380f6ec4968
it was split up like this only because the old worker system had no
promises. since we have them now we can use them for encapsulation.
Change-Id: Idbd523eff617d4c53c14b125ab3dfce4979cdab1
currently all loggers can always accept messages and never suggest
flushing buffers. in the future this may change, and at that point
we're already fully set up for it. local loggers should never keep
asynchronous (i.e. network-backed) buffers, disk buffers are fine.
networked loggers will require buffers and periodic flushes later.
Change-Id: Ide2114f5bc17f4a1d289c92ed4f9381a1d59dacf
this keeps the call stack flat even for very deep store directories.
moving progress reporting out of the method optimizing one path also
lets us move the inevitable IO wait and associated promise overhead.
Change-Id: I5083d88b05c247c7e19a21a646bcf3f62bfa3200
create activities from loggers themselves instead of passing the logger
as a constructor argument and allow direct construction of children, no
direct logger access needed. most call sites are not changed because we
still need to handle the "no parent" case, and the logger method can do
that more cleanly than a ternary at each site that creates an activity.
we may eventually want to create a root activity, which is cleaner too.
Change-Id: I295e056228dabb08a1316eba7973874784baa113
remotes can generate the same activity ids we might generate locally,
especially if the remote is accessed over the network. in that case a
pid collision is possible, and since activity ids are sequential it's
very possible to create colliding activity ids on both sides as well.
Change-Id: Id58074a41f5f7a59171b52818d1fb5a1beb4bf40
activities are scopes, and there's no reason scopes should not be able
to move around. this also means we can *create* them elsewhere without
also having to box them in some way, making parent relationships clear
Change-Id: I5df036e12ebd8270feb4dca1f23b6bee1f08e906
Currenlty `lix` and `nixos-module` have slightly diverging version
numbering. This commit fixes the lix side of things, in accordance to
the format proposed in
https://git.lix.systems/lix-project/lix/issues/585#issuecomment-7336
which allows for better use of `builtins.compareVersions`
fixes#585
Change-Id: I2072b701f9fadf780e7aab67b62fbcd2ec5df76d
this has no performance impact in any benchmarks we've run. nul bytes
are still used as implicit truncation points in many places all over:
rejecting them in all locations that treat them as a string end point
requires large changes such as using a proper path library everywhere
Change-Id: I936158bd435f6abf009a689adfbc24496262c578
This reverts commit 42691f0d94.
Reason for revert: darwin builds time out now, apparently due to newly appearing infinite loops in the test suite
Change-Id: I81a09efd53636a287f3095ac87ed8854b6e57922
The functional2 test suite was broken in the development shell for two mostly
separate reasons leading to "no such file or directory" errors:
* The `BUILD_TEST_SHELL` (already containing the correct path) would always be
set from the Meson option `build_test_shell`, even to the empty string if
that option is not set. Skip the overwrite in this case to make the
environment variable work again.
* Sandboxed builds would fail to find their builder `/bin/sh`, since in the
development shell busybox is not found, so no sandbox shell gets configured.
Use the shell from `BUILD_TEST_SHELL` instead. (How this does not break the
old functional tests remains a mystery.)
Change-Id: I6a6a696424e8caaef3f9b68e3738bfd58ea0b056
Local store make use of SQLite as a cache for output maps, they can
diverge with actual store reality.
Causing a crash at realization time.
We can just throw an error instead.
Change-Id: I943a4bddfd2461a34933e9ec079a6638ab58e5b1
Signed-off-by: Raito Bezarius <raito@lix.systems>
Fixes fj#940.
When running `nix-shell`, the `$NIX_BUILD_TOP` environment variable is
set to `$TMPDIR` or `/tmp`.
nixpkgs stdenv uses $NIX_BUILD_TOP to create `$NIX_BUILD_TOP/env-vars`
which contains all the environment variables set by stdenv. This is used
for debugging purposes in combination with `--keep-failed` to reload the
bash environment of a derivation.
`$TMPDIR` is often unset, therefore, `/tmp/env-vars` was constantly
being created. On a multi-user system or, when you run Lix as root, you
might create a `/tmp/env-vars` with different permission bits.
As a result, `nix-shell` can cease to function because that file will
fail creation for an unprivileged user for example.
fj#940 rightfully remark that the code is not consistent between
nix3-develop and nix-shell and it should be reworked.
Change-Id: Iddf15945385d8bd497b2800b37fee5e1f97689b7
Signed-off-by: Raito Bezarius <raito@lix.systems>
This avoids mangling debug gdb scripts and breaking them resulting in no
script loading at all.
Change-Id: Ib5e7034c37a1e41d1a1c9bd8442b8b8b8758e350
Signed-off-by: Raito Bezarius <raito@lix.systems>
UnsafeValueList held pointers from Value which were not necessarily GC
allocated, causing mayhem when evaluating something with genericClosure
(texlive environments).
We get rid of storing pointers and we let comparison take places on
const references.
We keep `gc_allocator<Value>` so that GC can scan things it allocates
inside the `res` list.
Co-authored-by: eldritch horrors <pennae@lix.systems>
Reported-by: qbit
Change-Id: I4fed3a3d9e18a2ef2d751f32d81801540e196f92
Signed-off-by: Raito Bezarius <raito@lix.systems>
thunk values are shareable, and we can represent invalid/uninitialized
values with a special bit pattern that makes no sense otherwise. there
is no need to keep allocating values on the heap, instead we can treat
values like reference-counted smart pointers to heap objects, which in
turn lets us save a lot of allocations and, ultimately, gc heap space.
compared to our baseline (main of 2025-09-27) we save 15%+ memory on a
system rebuild and 17% on nix search. eval time regresses by ~3% for a
system rebuild, while nix search is 7% faster. further optimization is
probably possible (but for now this will just have to be good enough).
Change-Id: Ib6c47acdbe2fac4f76a83c2269f16f30ef66b2e1
with thunk state being shareable we no longer need to worry about value
uniqueness, only about value lifetime. this means we can liberally drop
indirections and allocations, passing references instead of pointers or
using stack memory instead of gc-managed memory for some intermediates.
Change-Id: I2d48a6fd57a376d544bd9bd2d05e5420611986d1
there's no measurable performance gain in real-world testing to abusing
our list value storage like this. we haven't tested how much storage we
actually need on the stack to offset most of the temporary storage cost
and used 128 as a "good enough" value instead, reserving 1 kiB of stack
space on 64 bit platforms in a leaf function. this should do for a bit.
Change-Id: Ie98519b1da3e6fe685da88d1c44ffb4580fb592a
Value is already tagged. Value::Acb blocks are allocated for lambdas (so
we can fit the value tag into the three bits we have available), but the
current layout is rather wasteful for this purpose. the type bits can be
stored together with parts of pointers, which in the lambda case will be
the scope the lambda captures. the expr could also be used, but Env is a
gc-allocated item and thus guaranteed to be aligned properly for tagging
Change-Id: Ia685875387c7795bc4a00d73d1ce3cfea84e7297
this is a strong prerequisite for making values themselves copyable
without duplicating evaluation side effects. with this we can treat
`Value` the way we treated `Value *` to date and drop indirections.
Change-Id: I08f30d12697614a3ae7149615f6f1da83b13f52b
preserve type information of wrapped exception when constructing a
ForeignException. this requires using `std::current_exception` for
construction in all cases as passing the exception by reference to
any function erases dynamic type information: `throw` may copy the
thrown exception from its source expression; this only works as we
need it to if static and dynamic type of the expression are equal.
fixes#1000
Change-Id: I7167f298540de7c43c2566592bdb8d7a93c34d31
same as for null: we have few of them, they're statically allocated,
and they're not the largest contributor to the value population. not
storing them in Value itself frees up resources we *will* use later.
Change-Id: I521f9f243f48f56a78f7bffdf1dc1f0bc40a5c2d
we only need the one object for it. there's no need to waste precious
bits of the value internal type enum for this simple singleton datum.
Change-Id: Ie314b5bf429015e518798d9d65ad8ab2bb84a38e
floats are used very rarely, and our float support is bad enough to
strongly discourage using them on reproducibility grounds alone. we
can thus move them to more expensive storage without hurting folks.
Change-Id: I1086f612f85e294dd3fae4a2d334e09f52bbe4a8
external values very rarely appear during eval "normal" eval, and
creating them is pretty expensive. does *anything* even use them?
Change-Id: Id50fa3f76b7e1f551d550d99996a1ed5880b2531
despite not using allocation caches this does not have a statistically
significant performance impact, with less than 1% extra memory needed.
Change-Id: Ibe51a55ba986e471f217f3724977af17880fafff
using the same nodes as tApp is possible, and thanks to multi-arg app
nodes it can even be a bit faster than the linked lists used to date.
Change-Id: Idccb7c0b54c808e62da85d1c42ee09e6e92c4f7b
these behave like the old chains of app nodes, but they can store more
than one argument per node. for tApp values themselves this is not all
that useful, but if we could share tApp and tPrimOpApp backing storage
we could avoid creating and traversing the linked lists of values that
are currently needed to represent partially applied builtin functions.
Change-Id: I5a2a02d9733e1e0be5443459e2998d62fd3b9a5b
paths already are just strings with slightly magic semantics. the type
can mirror that at no perf cost, letting us drop one internalType tag.
Change-Id: I98acaa4fe3bedd28fc0841e1b81184d8dcddacc5
memory overhead is minimal and performance impact not measurable. once
we've done something like this for all value types that don't fit in a
single machine word we can cut a word from Value, offsetting the cost.
Change-Id: I9813bacd7e851957ad3426aed8f74033179a4212
we now use a single cache set for a number of sizes from one to eight
words. this also matches small attrsets, but perf impact seems small.
Change-Id: Icf16b329b98a20fcc9fe75e6395e148f0852c798
on its own this is not very useful, but having accessors for every value
kind is a prerequisite for doing smart things with Value than the union.
the net effect for now is only to add a few parentheses across the tree.
Change-Id: I88688ac09eb08495dad1eb221034ca540f094950
nix develop should ignore output checks in general.
This was done only for the old way of specifying output checks, the
structured attrs way requires rewriting the JSON and removing the output
checks pieces.
We take a brutal approach of removing as many as possible including
non-recommended ways of doing it.
Fixes#997.
Change-Id: Iaf83029016c71b5171e56e15d4eadc1a60a8be98
Signed-off-by: Raito Bezarius <raito@lix.systems>
In 7b37d5ea6a, aside from subdaemons getting
properly executed, they also lost the ability to outlive their parent, due to
now getting set the parent death signal like most other processes spawned by
Lix. This has annoying consequences like all concurrent builds being forcefully
terminated on system updates requiring a nix-daemon restart. As the behaviour
change was not documented and the systemd service file retained
`KillMode=process`, it seems to have been accidental. Restore the old behaviour
of letting the subdaemons outlive their parent.
Change-Id: I6a6a69645312a90dbce55495c2fef3825dd3c097
In preparations for a new representation of bindings that will make it
impossible to write an efficient `Bindings::find`.
Change-Id: I4e5a25b8d37d01b5728f7fe43978ceda2ab1b9b6
Signed-off-by: Raito Bezarius <raito@lix.systems>
Co-authored-by: Sergei Zimmerman <sergei@zimmerman.foo>
The way zipAttrsWith works is to replace the attribute set value by a
call to a function (the argument of zipAttrsWith) over the list of
attributes sharing a common key.
Instead of that, we will insert into the resulting attribute set the
various lazy calls and return that.
Change-Id: I2aae054eb99b1d1f8b0e7c658cc8d3488e5cdb01
Signed-off-by: Raito Bezarius <raito@lix.systems>
C++ has the "spaceship" operator which can be auto-implemented and
generates efficiently a strong ordering.
Change-Id: Idfd1fd68039b395e54401cbe913454e0cbd80fb3
Signed-off-by: Raito Bezarius <raito@lix.systems>
Prior to this change, references or pointers could be mutated. In
practice, we do not require this capability in the codebase except in
zipAttrsWith.
This cleans up all easy sites in preparation to have a smarter
representation of attribute sets albeit one that requires constant
references.
Change-Id: I2be20cce040a9228bde9e5f7b42c0499fba9550b
Signed-off-by: Raito Bezarius <raito@lix.systems>
Co-authored-by: Sergei Zimmerman <sergei@zimmerman.foo>
The stdio stream identifiers (stdin, stdout, stderr) are allowed to be macros.
In musl libc they are, for example doing `#define stdout (stdout)`, breaking
compilation with an error when one of the clashing variables is attempted to be
initialized the "wrong" way:
../lix/libutil/processes.cc:272:7: error: expected class member or base class name
272 | , stdout(stdout ? std::make_unique<AsyncFdIoStream>(std::move(stdout)) : nullptr)
| ^
/nix/store/ziw42d7rvgnf3vkbfc8kry07kipwf1xm-musl-static-x86_64-unknown-linux-musl-1.2.5-dev/include/stdio.h:67:16: note: expanded from macro 'stdout'
67 | #define stdout (stdout)
| ^
Other places only cause warnings on musl:
../lix/libutil/processes.cc:254:17: warning: parentheses were disambiguated as redundant parentheses around declaration of variable named 'stdout' [-Wvexing-parse]
254 | std::string stdout;
| ^~~~~~
/nix/store/ziw42d7rvgnf3vkbfc8kry07kipwf1xm-musl-static-x86_64-unknown-linux-musl-1.2.5-dev/include/stdio.h:67:16: note: expanded from macro 'stdout'
67 | #define stdout (stdout)
| ^~~~~~~~
../lix/libutil/processes.cc:254:17: note: add a variable name to declare a 'std::string' (aka 'basic_string<char>') initialized with 'stdout'
254 | std::string stdout;
| ^
| varname
/nix/store/ziw42d7rvgnf3vkbfc8kry07kipwf1xm-musl-static-x86_64-unknown-linux-musl-1.2.5-dev/include/stdio.h:67:16: note: expanded from macro 'stdout'
67 | #define stdout (stdout)
| ^
../lix/libutil/processes.cc:254:5: note: add enclosing parentheses to perform a function-style cast
254 | std::string stdout;
| ^
| ( )
../lix/libutil/processes.cc:254:17: note: remove parentheses to silence this warning
254 | std::string stdout;
| ^
/nix/store/ziw42d7rvgnf3vkbfc8kry07kipwf1xm-musl-static-x86_64-unknown-linux-musl-1.2.5-dev/include/stdio.h:67:16: note: expanded from macro 'stdout'
67 | #define stdout (stdout)
| ^
However they are still wrong, since the macro could be more complicated. Fix
them as well.
Change-Id: I6a6a6964a50ef7dec8f05f0bd8fc8f13f3036d51
It was supposed to be removed in 480fdf146d, as
it is not needed any more with the prelinked library. Due to a mistake in
rebase conflict resolution it reappeared by accident. Actually remove it now.
Change-Id: I6a6a6964d175fdb0ba0ad9ac55d4d22d7b27ad3f
Most tests for builtins now have `builtins.builtinName` as their name.
This makes navigating the test list a bit easier
Change-Id: Ief5af5c568a419bf9130601f9590e7a696b0dc0a
In the first pass I erred on the cautious side, only migrating safe
bets, to here's some trivial migrations that I missed
Change-Id: I934011919837b0aa491113afdcad603cf6b9cbbb
Closes#987
The patch adds a flag `--no-instantiate` which only performs evaluation
without instantiating any derivations. Hence, GC root creation is also
skipped. To achieve that, Lix is also put in read-only mode and all
operations that require reading a derivation (e.g. constituents or
listing input derivations) are disabled fallback values are set.
This is a port of an upstream PR[1]. Given the divergence of the
codebases (different restructurings on both ends, no more CA derivations)
I decided to redo large portions from scratch instead of
cherry-picking the patches. Hence, the authorship.
Additionally the clean up of casts down to a local store are removed or
guarded behind an if, as done in the upstream PR.
[1] https://github.com/nix-community/nix-eval-jobs/pull/379
Co-authored-by: Jörg Thalheim <joerg@thalheim.io>
Change-Id: Ib84f44e7799bc5577fd2ee98912458f16ebeab81
95448347 made lix require libatomic if the platform is able to link a
simple program using atomics, but it should actually be the other way
around. We need to require libatomic if it fails.
Change-Id: I6a6a6964ca6ee90a59314ddf1865753e83713772
using sleep(1) as a synchronization mechanism does not work. use fifos instead.
fixes#690 and ci constantly falling over in the same exact fucking source line
Change-Id: I51725f8e439b6753f3212d2897dbb0620ad77a37
It's pretty bad that Bindings effectively wasted whole 8
bytes (4 for capacity and + 4 for alignment padding) to
store something it doesn't actually need. BindingsBuilder
allows the capacity to be checked at construction time,
after which the Bindings does not get mutated aside from
the ugly case of builtins, which doesn't get built all
at once.
For `nix search --no-eval-cache github:nixos/nixpkgs/e1fa12d4f6c6fe19ccb59cac54b5b3f25e160870 hello`
this shaves off around 53MB allocations out of 2GB used
for attrsets in total:
< "bytes": 2001170768,
---
> "bytes": 1947398072,
< "Bindings": 16,
---
> "Bindings": 8,
Nix PR: https://github.com/NixOS/nix/pull/13919
Change-Id: I939c5ac545f5abbca048370dcf4936346339d75c
Some platforms like 32-Bit PowerPC need linking against libatomic.
Try to compile and link a very simple snippet of code which uses atomics
and make libatomic required if it fails.
Because we're using `dependency('atomic')`, the required meson versions
gets bumped to 1.7.0. See https://mesonbuild.com/Dependencies.html#atomic-stdatomic
Change-Id: I6a6a696471e1d352fb161c537ba9023b97c2d31e
libarchive is not async and cannot be used async without involving green
threads, which have already proven to be very problematic. unpacking tar
archives is rare enough that spawning a new thread for each shouldn't be
too much overhead, and the additional data copy probably also won't hurt
performance too much. we may even benefit from being able to extract not
just one archive per event loop but as many archives as we can keep fed.
Change-Id: Iece82bd566ada0a2a49de54c4e69caf6d93f6720
using a sink for this has long been a bit weird anyway. originally it
was necessary due to api limitations, but it hasn't been for a while.
Change-Id: I3dfa157944618349bfd6f398ee1667fc31519d86
Without https://github.com/NixOS/nixpkgs/pull/434761 evaluation of the
`nixpkgsLibTests` will fail in CI with recent enough Lix, due to reliance on
the TOML integer saturation bug.
Reported-by: Sergei Zimmerman <sergei@zimmerman.foo>
Change-Id: I6a6a6964838009d2c525f67035f84072fdfad988
Fixes: https://git.lix.systems/lix-project/lix/issues/973
Information about which commands were executed is really valuable to
debug Lix and is much more user relevant than the vast majority of the
e.g. build loop junk printed at debug level. Currently we have a *whole
lot* of call sites where we call execv* which should probably be cleaned
up, but that's future work.
I chose to print argv0 rather than the executable path if these differ,
since the code is shorter and since the command could be a fully
resolved symlink or so where argv0 is the actual command name being run.
However, it's not exactly *hard* to write std::ranges::drop_view(args,
1).
Change-Id: I73c3abb20b229d5e2d64277aa29cbbeed7764bab
printTaggedWarning already colorized its messages. we can do the same
for most other log messages.
Change-Id: Idcd31bbf4f8d0d703395b0d2b7b9bc33264d969f
luckily none of these a format strings vulnerabilities because
boost::format is smart enough to throw an exception when given
fewer format string arguments than are requested by specifiers
Change-Id: I5fa78f0d1396263271f6e1dbcee9c0b2e9e18c34
always use log macros, which also have the benefit of respecting the
verbosity setting without needing virtual function calls to read it.
Change-Id: I1c605562a53e54140724d5225e040abcf49ac996
we add two variants: one that just prints a message at the warning
level (mirroring the other printer macros), and one that also adds
the colored "warning: " prefix the function added. since there are
no overriders of this function in tree it looks safe to remove it.
Change-Id: I7008fd0f31d59fbc9259472e29359c8df19ff87d
mostly useful for nix-eval-jobs which currently has to call the logger
functions directly because its main code *isn't* in the nix namespace.
Change-Id: Ia8440d86a293d9006ffef2562b1859e9aaa79a62
Previously two cryptography libraries were linked into Lix: OpenSSL used for
hashing and (in usual configurations) indirectly via curl for TLS, and Sodium
used only for handling the Ed25519 path info signatures. The latter is
functionally redundant since OpenSSL supports the same use case as well.
Reimplement the Ed25519 handling using OpenSSL and drop Sodium.
Fixes: https://git.lix.systems/lix-project/lix/issues/969
Change-Id: I6a6a696456b9d3ad7fdc2bf9b0759836a6247a38
Currently, DerivationGoal prints a pretty generic message.
For many valid reasons, children may have better knowledge of the detail
of what has happened and would like to extend the error message.
What we did is to printError at convenient places but this is
counterproductive because the build error can bury the notes.
This is still not perfect because there's no fine-grained structured
information that children can use to act upon the generic messaging, but
this is already an improvement for LocalDerivationGoal and keep failed
which will occur in the next change.
Change-Id: I5835cbbb30c4f2aa64abefb83999018d30ca4a0c
Signed-off-by: Raito Bezarius <raito@lix.systems>
It was only used for impure derivations, which were finally removed in commit
be07629820. Delete the unused function.
Change-Id: I6a6a696481711f68a8c3ea7eac7978fcf5884cce
Closes#551
This adds a special accessor that falls back to checking if a store-path
exists within a chroot if it's not a valid path. That way,
`genGraphString` can find out which files have which references before
the outputs are registered.
Change-Id: I03c9d508fa3c72e5c262194461a25d71f3f4de15
That way it's possible to inherit from LocalStoreAccessor to implement
special behavior such as an accessor that falls back to the chroot
directory if it can't find a store-path (which is what we'll do in the
next commit).
Change-Id: If689eb3f410e81e629f1d13cc2b48594fecb1001
That way we do now have linear complexity to determine output
references per output within each step of the topological sort.
Instead, this is done before and the topo-sort only filters the output
map for other derivation outputs.
Following up on this, we can re-use `outputGraph` to generate a tree
with references to display which files cause an output reference cycle
if needed.
Change-Id: Ibdd46e7b2e895bfeeebc173046d1297b41998181
If none is given, we fall back to whatever accessor we get from the
store.
To display which paths actually contain the references leading to
e.g. a cycle or triggering a disallowedRequisites error, we'd
potentially have to look into the chroot from the previously finished
build. This behavior should not be part of the local accessor by
default, but part of a "special" accessor. This change allows using such
an accessor for `genGraphString()`.
Now that we inject the accessor from the outside, we have to mock it
anyways in the tests. Hence, this also adds a testcase for the
precise=True case.
Change-Id: I58465fb944776c2b0262ba054d1f296ed2ae3406
The variant has on the left-hand side the topologically sorted vector
and the right-hand side is a pair showing the path and its parent that
represent a cycle in the graph making the sort impossible.
The goal is to implement #551 which needs to throw an error if the
topo-sort fails. However, the error-message is supposed to contain a
graph of store-paths and the API to generate this is inherently async.
Now, catching the exception and re-throwing another one is impossible
since `co_await` is forbidden in `catch`-blocks and adding another
topoSort variant that allows an async `makeError` also seems odd. Hence,
I decided to alter the data-structure in use a bit for this use-case.
One out of two uses of the function are affected after all.
Change-Id: I70a987f470437df8beb3b1cc203ff88701d0aa1b
it's broken, can write arbitrary file paths when run as root, and only
supports strings and recursive sets of strings. this was only used for
manpage generation in a build system that has not woken up since 1976.
fixes#974fixes#227
Change-Id: I4f18599685a3077c15ddc02c759558f986c8c6e4
Commit 5dc847b47b introduced it as a non-inline
function with definition in the header, which can result in linker errors like
the following:
/build/source/build/lix/libutil/backoff.hh:36: multiple definition of `nix::backoffTimeouts(unsigned int, std::chrono::duration<long, std::ratio<1l, 1000l> >, std::chrono::duration<long, std::ratio<1l, 1000l> >, std::chrono::duration<long, std::ratio<1l, 1000l> >)'; tests/unit/liblixutil-tests.p/libutil_backoff.cc.o:/nix/store/eeeeeeeeeeeeeeeeeeeeeeeeeeeeeeee-gtest-static-x86_64-unknown-linux-musl-1.17.0-dev/include/gtest/gtest-printers.h:1223: first defined here
This error was observed during trying to bump `lixPackageSets.git` in nixpkgs.
I am not sure why it can't be observed in the in-tree `nixStatic` package but
the definition is wrong in any case.
Change-Id: I6a6a6964e218a03ca2a2e8eddbb72d44e06e904e
Closes#932
`connect-timeout` gets replaced by an exponential backoff for the
download timeout where the initial value is controlled by the setting
`initial-connect-timeout`.
Per iteration, the upper limit of the timeout is increased set to
timeout := min(max_connect_timeout, initial_connect_timeout * 2^i)
I decided to move the entire timeout / tracking of attempts into its own
class to not make the filetransfer implementation more complex. Also,
that allows us to write unit-tests for it.
Setting `--download-attempts` to `0` is forbidden now and an exception
will be thrown. For `--offline` we set it to `1`, the behavior is
equivalent to what it was before: whether the max tries were exceeded is
only checked after the first download exception got thrown, i.e. there's
still one attempt being made.
The end-result - with timeouts being caused by a wrongly set proxy -
looks like this:
$ env HTTPS_PROXY=1.1.1.1 nix store ping --store https://example.com
warning: error: unable to download 'https://example.com/nix-cache-info': Connection timed out after 5006 milliseconds (curl error code=28); retrying in 422ms ms (attempt 1/5)
warning: error: unable to download 'https://example.com/nix-cache-info': Connection timed out after 10010 milliseconds (curl error code=28); retrying in 1003ms ms (attempt 2/5)
warning: error: unable to download 'https://example.com/nix-cache-info': Connection timed out after 20020 milliseconds (curl error code=28); retrying in 2018ms ms (attempt 3/5)
warning: error: unable to download 'https://example.com/nix-cache-info': Connection timed out after 40007 milliseconds (curl error code=28); retrying in 4087ms ms (attempt 4/5)
error: unable to download 'https://example.com/nix-cache-info': Connection timed out after 80074 milliseconds (curl error code=28)
Change-Id: I9e8d08d78275bcf60080d663febc9e075243d36b
Currenlty, when a test group is invalid already, we also throw unsued
file errors.
This leads to clutter as more often than not, the unused files are
caused by an invalid configuration, making the debug stack bigger
without reason.
With this commit the behavior is changed to only error about unused
files, when no other configuration issues were found
Change-Id: I92a819753f13b8ed5a07dae53ecaee5d84b5ce64
Currenlty one is required to always write the bulky `mark.parametrize`
with indirect and things
This commit adds a custom decorator for usage of files, which hides the
parametrization complexity from the user.
Change-Id: I526e016d12006669dc302dfc5af619735399c503
Concept: what if you could, in your fancy terminal, in the year of our
lord 2025, just click on the attrs you're looking at to go to where
they're defined. Currently we only expose this info as
builtins.unsafeGetAttrPos, which is inconvenient as it's not
discoverable to users.
By putting it in this more visible yet invisible spot, it's more likely
to be more useful to more people.
In the current state, this is not the most useful ever due to stuff like
https://github.com/neovim/neovim/discussions/35097. However, it can be
expanded by perhaps adding something like the url format setting ripgrep
has.
Change-Id: I3947f97d5c2056d59099af468d7b855486438227
this pretty much only impacts store verification via the nix3 cli. no
other thread pools are left, and the verification pool may *actually*
be important for throughput since verification involves much hashing.
Change-Id: I32152e6169a82a1268a790e333f21a0430ede7f4
signing is very cheap, it's only the store access that is expensive.
http binary caches parallelize async accesses extremely well though.
Change-Id: Ifdbf398bd328ba16ec4e8caba3f5f99a6cf3e046
this is used by nix-env and copyPaths, which in turn is used to upload
to binary caches. for a large path set we have seen 10x a improvement.
Change-Id: Ieadd0e66180e5ceecefaf944a5bb2f0523374954
currenlty we use the external package `toml`, this just adds an
unnessecary dependency, as python ships its own toml as `tomllib`
Change-Id: Ia63fa7558973e853ada20cbfa21d897d700444f8
Always use tmpDir on darwin. Call setupConfiguredCertificateAuthority even if useChroot = 1 on non-linux.
Even though macOS builds are not executed in a chroot, enabling the sandbox
sets `useChroot = 1`. Basically, useChroot is set when the sandbox is enabled,
not really when a chroot build is executed
Change-Id: I8d4c1e617abcc05dfabd998a8ce94bb11587f9d1
We always use the default temporary directory, because
`createUniqueDir` has an interface nice enough to use directly for
the few bespoke uses.
Change-Id: I6a6a6964c15c31bb3e131fbe1db1837987a6d6dc
This makes the paths more nondeterministic, but more reliably unique,
and lets us remove the retry loop.
Note that this adds random entropy to the build directory visible
inside derivations on Darwin and unsandboxed Linux. It was already
non‐deterministic in the presence of concurrent builds and similar,
but now we can reliably expect it to be different every time. On the
whole I think that’s a good thing, as it is impossible to ensure
a single consistent build directory and derivation outputs should
not depend on it.
Package reproducibility isn’t great on Darwin to begin with,
though, and the reproducibility bugs this will turn up in packages
will be more urgent to fix than when the build directory was mostly
consistent. A quick survey of my local store shows that many C, C++,
and Rust binaries contain build directory references, likely due to
use of `__FILE__` and its equivalents; non‐binary offenders include:
* Install logs included in the Rust and Cargo bootstrap compilers
* Example errors in the Rust documentation referencing build paths
* Configuration information installed with CPython itself
* Python 2 metadata from resholve’s closure
* Cython metadata
* Generated headers in Facebook libraries referencing source paths
* Generated CMake files in Facebook libraries referencing source paths
I haven’t built that much in this store since the last GC, so this is
probably only a small sample of the problems across the tree. These are
all instances of <https://reproducible-builds.org/docs/build-path/>,
though, and should probably just be treated as general reproducibility
bugs outside of contexts like the Linux sandbox where we can normalize
them away entirely.
I have implemented away build directory paths for C/C++, applied some
additional fixes for non‐`__FILE__`‐related issues in binaries
from ATF and LLVM, and fixed the derivation bug causing the CPython
3 issue, and will work on upstreaming these changes. Rust is working
on the problem upstream, with some temporary workarounds we can
potentially apply in Nixpkgs for now. The rest will require some
distributed effort.
Change-Id: I6a6a69648f74d85c6fca86cc52f38fd957e4f9ad
This does change the behaviour when the global temporary directory
does not exist, but other uses of the global temporary directory are
already broken in that circumstance, and it should be fixed centrally
if the use case is considered desirable. The logic was not present
before the recent churn around build directories – it was added now
that Lix is taking ownership of the build directory in the store –
so this should not be a meaningful regression.
Change-Id: I6a6a69648054ae201b3ce36d11e49c93793fdb0e
There is now no risk of race conditions on a system with a functioning
entropy source, and the bespoke prefixes are either redundant to the
default or unnecessary.
Change-Id: I6a6a69648a3b8060333e97269ea8b72499614559
Relax the constraints on keeping the exact same filename format to
provide a more robust source of entropy with a simpler interface
(as previously suggested by eldritch horrors). Using 128 bits of
OS‐provided entropy ensures global uniqueness and allows us to
skip any thought of gracefully handling the case where these files
already exist.
My microbenchmark that repeatedly constructed paths like this and
printed them out showed that this takes about 1.23× the time of
the previous implementation, both taking on the order of a couple
microseconds for one iteration. Since everything that uses it is doing
things more expensive than printing to standard output, the actual
performance delta is likely to be lost in the noise. If it somehow
becomes a bottleneck, it can be optimized without sacrificing the
guarantees by reading from the system RNG only to seed a thread‐local
CSPRNG like [ChaCha8Rand], but I think that’s very unlikely.
We also tweak the recommended way of creating a temporary file inside
a directory in anticipation of later changes, and rename the `suffix`
parameter to `prefix` (it’s a prefix to the random characters and
a suffix to the root, but this way is more consistent).
[ChaCha8Rand]: https://c2sp.org/chacha8rand
Change-Id: I5bd7badf1392243f485935c4a016c1f833cb16d3
The prospective callers of this should probably be doing something
smarter or more abstracted to begin with, but this is useful as an
incremental improvement for call sites with existing `makeTempPath`
logic in the face of filename length limits.
Change-Id: I6a6a6964374f47abbf0ec10aa8d945c4e50a43af
notably this also includes the symbol table because it stores real
strings that are referenced by eval values, and an upcoming change
will make it impossible to share those strings with value strings.
Change-Id: I20a3644db8aa0850efe29630e0b73d424cb2aa56
See [my comment] on the Nix PR to restore the previous behaviour
for why I believe we should remove this for the next release. The PR
should still be backported to stable releases to avoid making breaking
changes to their semantics.
[my comment]: <https://github.com/NixOS/nix/pull/13741#issuecomment-3180851635>
Fixing this across supported Lix versions is required for Nixpkgs to
update toml11, which is a blocker for the CMake 4 update.
Change-Id: I6a6a69642e6b6cb13a9fccc0778e9158b53102d5
This version changes the handling of TOML timestamps, and throws an
error on out‐of‐range integer literals rather than the previous
saturating behaviour, as required by [the TOML v1.0.0 specification]:
> Arbitrary 64-bit signed integers (from −2^63 to 2^63−1) should be
> accepted and handled losslessly. If an integer cannot be represented
> losslessly, an error must be thrown.
[the TOML v1.0.0 specification]: <https://toml.io/en/v1.0.0#integer>
The only known use of this is a questionable Nixpkgs test that I have
proposed [a fix] for.
[a fix]: <https://github.com/NixOS/nixpkgs/pull/433710>
Bumping this ahead of Nixpkgs ensures we can test these cases on
HEAD in advance. I presume that the next Lix major version will be
released after 25.05 goes out of support, so it should be fine to
drop support for the old version of toml11.
The co‐authors of this commit are the contributors to the vendored
package definition from Nixpkgs.
Co-authored-by: Anderson Torres <torres.anderson.85@protonmail.com>
Co-authored-by: Artturin <Artturin@artturin.com>
Co-authored-by: Silvan Mosberger <silvan.mosberger@moduscreate.com>
Change-Id: I6a6a69644a188b6e09eee5c9cf91ddd3c81d24ee
This behaviour is bad and will be fixed, but adding a test for the
present state documents the change.
Change-Id: I6a6a6964b88dc929269ee136804857b3852dfafe
This addresses several changes from toml11 4.0 bump in
nixpkgs [1].
1. Added more regression tests for timestamp formats.
Special attention needs to be paid to the precision
of the subsecond range for local-time. Prior versions select the closest
(upwards) multiple of 3 with a hard cap of 9 digits.
2. Normalize local datetime and offset datetime to always
use the uppercase separator `T`. This is actually the issue
surfaced in [2]. This canonicalization is basically a requirement
by (a certain reading) of rfc3339 section 5.6 [3].
3. If using toml11 >= 4.0 also keep the old behavior wrt
to the number of digits used for subsecond part of the local-time.
[1]: https://www.github.com/NixOS/nixpkgs/pull/331649
[2]: https://www.github.com/NixOS/nix/issues/11441
[3]: https://datatracker.ietf.org/doc/html/rfc3339
(cherry picked from commit dc769d72cb8ad22a0f89768682b5499a9d2b3d8b)
Upstream-PR: https://github.com/NixOS/nix/pull/13741
Change-Id: Iac4fbe5108be79be585e9670fa42dfd11f3c5e89
There's no reason to use a std::function for recursive lambdas
since there are polymorphic lambdas.
(cherry picked from commit a80a5c4dba0d944fab8f5ed57a343869ae96bf16)
Upstream-PR: https://github.com/NixOS/nix/pull/13741
Change-Id: I593bd04597e2ae000374ca1eca4d8928e986c0b5
This looks really weird after the reformat.
(cherry picked from commit df4e55ffc13c413e270af134227115a20a2341ba)
Upstream-PR: https://github.com/NixOS/nix/pull/13741
Change-Id: I8de92d58620cc4545a31d8b7d533d2f1e9f4f233
JSON has only one numeric type, with vague semantics. [RFC 8259] says:
> This specification allows implementations to set limits on the range
> and precision of numbers accepted. Since software that implements
> IEEE 754 binary64 (double precision) numbers [IEEE754] is generally
> available and widely used, good interoperability can be achieved by
> implementations that expect no more precision or range than these
> provide, in the sense that implementations will approximate JSON
> numbers within the expected precision. A JSON number such as 1E400
> or 3.141592653589793238462643383279 may indicate potential
> interoperability problems, since it suggests that the software that
> created it expects receiving software to have greater capabilities
> for numeric magnitude and precision than is widely available.
>
> Note that when such software is used, numbers that are integers and
> are in the range [-(2**53)+1, (2**53)-1] are interoperable in the
> sense that implementations will agree exactly on their numeric
> values.
[RFC 8259]: <https://www.rfc-editor.org/rfc/rfc8259.html#section-6>
Floating‐point numbers are annoying to deal with in Nix, so it
optimistically parses integer‐looking literals as Nix‐language
integers where possible. Nixpkgs relies on this behaviour, as it backs
its `lib.toInt` family of functions with `builtins.fromJSON` in lieu
of a real integer‐parsing built‐in, and treats floating‐point
outputs as an error. Therefore, dealing with integer‐looking
JSON number literals that are outside the interoperable range is
unavoidable.
However, this raises the question of how literals that look
like integers, but exceed the range of a Nix‐language integer,
should be handled. The JSON library we use attempts to represent
integer‐looking literals as a unsigned or signed C++ integer type
before falling back to floating‐point numbers.
This means that we were parsing literals below −2⁶³ as
floating‐point numbers, while rejecting ones above (2⁶³ + 1)
with an error. This was done to avoid the C++ undefined behaviour in
the previous code path, but is hard to justify. This change causes
them to both be parsed as floating‐point numbers.
The alternative would be to reject the signed case too. However,
I believe that is less consistent with how JSON handles its single
numeric type, less interoperable with JSON documents found in the wild,
and that it is preferable to avoid the function being needlessly
partial in this case. It does mean that round‐tripping is less
lossless than before, but extreme floating‐point values already
caused these kinds of issues, and from JSON’s point of view that’s
exactly what these literals represent. Only numeric values outside the
RFC’s suggested interoperable range will have any round‐tripping
issues, and we continue to guarantee the behaviour on all values that
can be represented as Nix‐language integers.
Change-Id: I6a6a696412383e8a2cc160397716cb7f7bc7a2d4
So far, the environment used by `command` was completely leaky and the
one used by `nix` was very leaky despite it trying to be a "hermetic"
environment.
This commit moves the hermaticity to `command` and changes its
implementation to be not leak anything.
To achieve this, the following changes were also nessecary:
- the `files` and `snapshot` fixture now use the folder `test-home`
within the tmp_path directory by default, as the `HOME` environment
variable is set to there. (extraction not possible due to dependencies
of command etc also using this directory)
Fixes: #847, #848
Change-Id: I55f86ee0e1615e73fcf442ee2f28f3b89893bbb4
Current test suite doesn't cover the subsecond formatting at
all and toml11 is quite finicky with that. We should at the very
least test its behavior to avoid silent breakages on updates.
(cherry picked from commit 7ed0229d1abd4414144c7af396842462ce6fc1eb)
Upstream-PR: https://github.com/NixOS/nix/pull/13741
Change-Id: I6a6a696433b168072d6ad2585dce8a3c10ccbc39
use a thread pool and allow only buffered sources for performance. we
may want to use this code path unconditionally due to gc interactions
of fibers in circumstances we should not even be able to trigger, but
if that becomes important we will have a useful implementation ready.
Change-Id: Ib4e1531fe920847d8e30a42e8df393ace549f52e
We need capnproto-lix to be provided by callPackage, otherwise it won't
be spliced and the same (built for the cross host platform) for both
buildInputs (correct) and nativeBuildInputs (incorrect). We thus move it
into the overlay.
A similar problem exists for the lowdown build. We thus use callPackage
to override it as well. This is horrible (especially because it means we
need to pass the enableDarwinSandbox package argument through
ourselves), but at least it builds...
Fixes: https://git.lix.systems/lix-project/lix/issues/939
Change-Id: I802152072d852903401ef701f526195aa99475f2
When testing specific internal functionallity while needing things from
the testlib, so far, the tests for the testlib have always been copied
too.
To reduce the amount of additional program required when later making
the env of the pytest_command declarative, and to not test the same
tests a multitude of times (and potentially reaching infinite recursion)
those tests will no longer be copied
Change-Id: I36ec3824a21ed30f9b8ff19948031d1edbf6c76c
this means both the worker protocol and the serve protocol, i.e.
ssh-ng/local connections and legacy ssh connections. now we have
no blocking reads left anywhere in our client store connections.
Change-Id: I2f628d4d2e71ef0a7006918f175192f3f58eea95
`mesonCheckPhase` of course runs `preCheck` and `postCheck` hooks,
not `preInstallCheck`, so this was not doing the right thing. Really
sorry for breaking CI :(
Fixes: e6da29ad6b
Change-Id: I6a6a69642a242e2c8bfe10ac61d3f6756d666548
Snix's nar-bridge[1] stores NARs under a different URL, i.e.
`nar/snix-castore/<hash>.nar` rather than `nar/<filehash>.nar`. Right
now, when copying into such a store via the http binary-cache, we'd end
up with wrong cache entries that point to the wrong NAR URL.
On Hydra, this is a fatal error, i.e. builds that depend on previously
built paths (that were written to the cache before by the queue runner)
would be aborted because of that.
This patch removes the caching since we'd have to re-fetch the narinfo
to do taht and this can also happen the next time, the narinfo is
queried. Also, removes the negative cache entry indicating that the
store-path doesn't exist in the store.
We don't have any coverage for http-stores so far, so I wrote a small
testcase for the "default" case and the nar-bridge case in functional2
since it has a very nice fixture for an HTTP server ready. I'm aware
that there's a CL for a nicer cache server[2], but given I'm down a
pretty deep rabbit hole by playing around with Snix, I decided to not
finish the CL and write something small for the tests in here. This can
be replaced by the fixtures from that CL later on as well.
[1] https://snix.dev/docs/components/overview/#nar-bridge
[2] https://gerrit.lix.systems/c/lix/+/2431/1
Change-Id: I4fcdf47a6bf9c3c8fbeb235eeca7a48914a4d693
shoving a nar dump directly into a framed sink created a bunch of tiny
frames, each of which requires at least two syscalls to read. this can
lead to immense performance loss when using a daemon; we have seen 15%
in benchmarks on main and even more with async code involved ... oops.
Change-Id: I8529506e3de74d92834d1f4ee228dcaf32eb756c
Notably, this adds the Requires.private required for proper static linkage of
Lix libraries. Some minor missing or duplicated dependencies are also fixed
along the way (although some optional dependencies are omitted due to their
size).
Fixes: https://git.lix.systems/lix-project/lix/issues/789
Change-Id: I6a6a696413d538124d9ac75c68f100cc3089284f
A bunch of dependencies were superfluous, while others were missing (in
particular on internal sub-libraries) and things just happened to work because
they leaked in or were explicitly required in other places. Make efforts to fix
them all.
Change-Id: I6a6a69643e35ac4b0b66a72f4e42a2ba5ed52488
Some of the Lix libraries always need to be linked in full due to their
reliance on static initializers. This was achieved internally using link_whole,
but they are still easy to abuse by external users who manually need to
remember passing linker flags such as `--whole-archive` (GNU) or `-force_load`
(Apple), and the obvious way to shove it in pkg-config breaks Meson due to
potentially including a library's flags multiple times, and then deduplicating
only the file names leaving a stale `-force_load` around causing trouble.
Instead we now "prelink" the static libraries, by merging them into one object
file. Since the static linker will always link entire object files, this will
have the same effect as whole-archive linking (except the library won't be
included if it's completely unused, which should not cause trouble since it's
unused after all, and dynamic libraries behave the same way). Unfortunately
Meson's native prelink functionality cannot be used due to missing (non-Apple)
Clang support [1], so write our own one. While not particularly portable, it
should work with Clang which is the only officially supported compiler, as well
as GCC.
[1] https://github.com/mesonbuild/meson/pull/14846
Change-Id: I6a6a6964a82241ce3b0b11fe8397fd451b8027f2
Meson only requires the `modules` option for Boost libraries that should be
linked against [1]. However, we use only header-only portions of the Boost
container library; in fact we only do not run into the disallowedReferences
because the linker drops the unused library. Remove the misleading option.
[1] https://mesonbuild.com/Dependencies.html#boost
Change-Id: I6a6a69648b806bf6bcf784391263f5ee9cd63a0b
This gets us `--print-errorlogs` and `--timeout-multiplier=0` for free,
and also fixes the concatenation logic with `__structuredAttrs`.
Change-Id: I6a6a69643ea7224cb24508505c554143f83ae243
Otherwise we'll get system names like `x86-linux`, which is nonesense.
Also add fixups for 32-bit ARM, and MIPS.
Change-Id: I6bd773bec82dcc54b40bbc018641cd614b272a28
this lets us avoid a bunch of fcntl calls to unset and re-set O_NONBLOCK
on remote store connections. the overhead of these isn't high, but doing
it is still wasteful and a maintenance burden when we have async readers
Change-Id: I900cdca2a16202380c8b6f9b86da7d9b0f1e34ac
the former isn't even int, it's unsigned. the latter is uint64. both
should be explicit about this to avoid problems that we have already
had, such as the FramedSource wire protocol writing using 64 bits to
write frame sizes, but 32 bits to read them. large frames will cause
the reading end to crash with an unnecessary serialization exception
Change-Id: I3c15e911f649eec719d4b1c135dde1b6ba020271
they will not work well with async deserialization and are not used
consistently anyway. just like the serializing operator<< these are
protocol stability hazards: changing the type of a field influences
the wire protocol layout and type constraints, which is not amazing
Change-Id: I54b20a133048f4ca15a9fb0f4d8b94dc78f62d89
this is an equivalent of the regular kj read interface which also takes
a min/max pair. we do not need this very often though, so we'll keep it
as a separate method for now. if we do find we use it more than read we
can still rename read to readSingle and readRange to read. we will see.
Change-Id: Ib04ca146911adae7081cf4b2df097217ea5fe9f8
this should've been a filter from the start. since filter support in the
old IO model is very bad we just move it into the only use of it instead
Change-Id: Ifb9cfecf07587ae1d2d55072ddf505c86c79cc1b
we should not encourage this kind of framing. in the future we will have
to do this on async streams, which we *absolutely* should not encourage.
Change-Id: Ib89e144afb564284db64fc7367cba7fffc18fdaf
the exception no longer actually matters, only that the download stream
is destroyed before the download finishes does. exceptions during drain
calls on the returned stream will cause the stream to be destroyed, but
draining this stream is just a remnant of the old download-to-sink API.
Change-Id: Ic6de40deb2ccff09d77180148afe746f29f55d23
Nixpkgs expects the `builtin.currentSystem` for POWER CPUs to be:
`powerpc[64][le]-linux`
But using `host_machine.cpu_family()` for the CPU part of the system
string on POWER produces this instead:
`ppc[64]-linux`
So evaluating Nixpkgs errors out on:
`error: Unknown CPU type: ppc64`
To fix this, change
`ppc` -> `powerpc`
`ppc64` -> `powerpc64`
and append `le` if `host_machine.endian() == 'little'`.
I can't actually test this on hardware rn due to hitting a kernel bug
on the host system when linking big things[1], but the approach here
is similar to how it was fixed in cppnix[2][3], so it *should* be fine.
[1] https://git.adelielinux.org/adelie/packages/-/issues/1315
[2] https://github.com/NixOS/nix/pull/13514
[3] https://github.com/NixOS/nix/pull/13520
Change-Id: Ib82839cdaf2198bf18b89e82caaa1217f88e11ed
This is a collection of Lix plugins that showcase how to write one for
various usecases.
The first is a mTLS store plugin that enable mTLS cache URIs
(`https+mtls://`).
We enable meson build system support for this plugin but we are not
going to distribute it in the official packaging of Lix, we will
repackage each relevant plugin downstream in Nixpkgs.
These plugins have *NO* guarantee support, they are provided as useful
references and are possibly production-ready if your usecase is simple
enough.
Reference: https://github.com/NixOS/nix/pull/13030 (this change has
resemblances but our APIs are different, the tests harness is mostly
from CppNix).
Change-Id: Ib354271981b35dff6c134b12c4748c3eaf743fcb
Co-authored-by: Jörg Thalheim <joerg@thalheim.io>
Co-authored-by: László Vaskó <1771332+vlaci@users.noreply.github.com>
Signed-off-by: Raito Bezarius <raito@lix.systems>
This is useful to test extended features in Lix.
Change-Id: Idb2416a080329116677809b883950e6c33028a44
Signed-off-by: Raito Bezarius <raito@lix.systems>
we don't need to report progress for every read call. that's way too
much. batching like this greatly reduces CPU usage for copies out of
or into remote buidlers due to likewise greatly reduced log traffic.
Change-Id: I3db2b2ab113fbaadefc69cfde6f977fb0c6cd5ad
Historically, Nix would support copying certificate authorities inside
the sandbox so you could use them.
In addition to that, the primitives consisting of leaking environment
variables via `impureEnvVars` and `extra-sandbox-paths` to render paths
external to the sandbox visible to the builder would also constitute a
mechanism to expose special inodes which should have no influence on the
output result, e.g. interception CAs.
Unfortunately, in nixpkgs, `lib.fetchers.proxyImpureEnvVars` set
`NIX_SSL_CERT_FILE` as an impure environment variable.
A confused user may set `ssl-cert-file` via `NIX_SSL_CERT_FILE` outside the
builder believing that this will set magically the right
`NIX_SSL_CERT_FILE` inside the sandbox, but this is not true.
The combination of impure environment variables and setting `caFile`
creates a weird interaction where `NIX_SSL_CERT_FILE` points to an
"outside the builder's world" inode *AND* `ssl-cert-file` creates
this very same certificate file in /etc/ssl/certs/ca-certificates.crt
without rewriting the environment variable.
This footgun is closed by making these two features mutually
incompatible with a warning and forcibly rewriting the SSL family of
environment variables even if it was set via impure environment
variables.
Users who truly meant to use `impureEnvVars` can obtain the right
behavior by setting `ssl-cert-file` to an empty string and will have to use
`extra-sandbox-paths`.
Users who meant to use `ssl-cert-file` will have everything work
automatically with a warning hinting at nixpkgs *fixing its own bug*,
i.e. passing `NIX_SSL_CERT_FILE` as an impure environment variable and
expecting the Nix interpreter to magically reconcile the diverging
values or expecting the user to actually do the work to render the path
visible available via `extra-sandbox-paths`.
Fixes#885.
Change-Id: I32f8b5ce20fe9b6a911768114c92f95fc886cc07
Signed-off-by: Raito Bezarius <raito@lix.systems>
Sometimes, `bindPath` will detect the source is a symlink and we are not
using the new mount API which support symlinks (kernel ≥ 5.12 IIRC?).
In those instances, we copy the inode to the target.
But some callers may want to follow the symlink in such circumstances,
we add a new default argument to the previous value and let caller
decide for themselves.
Change-Id: I8505b613fc614ce539eb89258fbbb7eaecebe23b
Signed-off-by: Raito Bezarius <raito@lix.systems>
`pathContentsGood` is used to assess the validity of a path as part of
derivation goals *in repair mode*.
When repair is used with a diverted store, i.e. a store where
fsPath(toRealPath(s)) != fsPath(s) for s a store path, this result in
utterly broken behavior because it will attempt to assess the goodness
of the *logical* store locations, most of the time: /nix/store/...
So, if you are repairing your system using a live NixOS ISO. Your ISO
contains a `/nix/store` (assumed to be good) and you repair your system
which is rooted at /mnt and contains its own /nix/store, that is, a Nix
store at /mnt/nix/store.
Performing the following operation `nix-store --verify --repair --store
/mnt` will assess the contents goodness of the ISO's Nix store.
To avoid this, we assess the path existence of the *physical path*, aka
the result of `store.toRealPath` applied to a *logical* store path
string representation and we verify the hash of the *physical path*.
The error messages are not taken care of in this CL as those are purely
cosmetic and helps the user understand what is going on.
Fixes#892.
Change-Id: Ib9e0153cb5683edcf37f1963ebf065ceba5e5dfb
Signed-off-by: Raito Bezarius <raito@lix.systems>
Instead of showing logical Nix store paths, we show the actual physical
location so that the user can stat by copy-pasting these paths.
The bad thing about this change is that certain Nix porcelain only
accept their logical counterparts.
Change-Id: Id0eb45d1bf08a23508dfc2bb694c88155654f585
Signed-off-by: Raito Bezarius <raito@lix.systems>
When a user runs a repair or check sequence, they might be confused of
seeing references to a logical /nix/store path rather than the actual
physical store location.
Change-Id: I042031a6159cdd1054e7e200a220bf6c321d5fb8
Signed-off-by: Raito Bezarius <raito@lix.systems>
jade: I rewrote this PR to be consistent between nix-instantiate and nix
eval. It turns out that nix eval *doesn't* copy to store with `--json`,
whereas nix-instantiate does. Wat.
Closes: https://github.com/lix-project/lix/pull/17
Co-authored-by: tyberiusprime <tyberiusprime@noreply.git.lix.systems>
Change-Id: Id22deec1cee0fed3bd5689567869b70bab26bae5
Originally, libboost_context and dependent dynamic libraries have been copied
into the nix package to remove boost from the closure and consequently reduce
the closure size. Since commit ef0de7c79f we do
not depend on Boost coroutines any more, so these libraries are not needed at
all any more and (somewhat ironically) only increase the package size. Remove
them.
Change-Id: I6a6a6964dc3e0b29dfad8b2b232b428ba3cc653c
Meson conveniently does lets you pass feature objects to `required :`
arguments[1], which is handy
[1]: https://mesonbuild.com/Build-options.html#features
Change-Id: I54194b235a9b3dc207f3f78e0a8c50f957e1fd1f
The issue requiring these environment variables to be set for Meson to find
Boost [1] has been fixed [2] for quite some time now. Drop them since they are
unnecessary in all supported nixpkgs releases (in fact they have already been
removed in the lix package there).
[1] https://github.com/NixOS/nixpkgs/issues/86131
[2] https://github.com/NixOS/nixpkgs/pull/315998
Change-Id: I6a6a69640a30e917cd7a983b9d846d023b393dcd
In certain scenarios, a path may exist but is a broken symlink. For
instance, this happens frequently if you are rescuing an existing NixOS
system via `nixos-enter` or a manual `chroot` invocation because the
rescued system `/etc` may have broken links or the system prior to the
root pivot may interfere.
Nonetheless, these dangling symlinks are not always fatal for the builds
so we can just warn and skip their setup.
The warnings will provide a better diagnostics for system administrators
also.
Fixes#893.
Change-Id: Ifa12be3a43f23c973d7b466e8b73bd776abf3e7b
Signed-off-by: Raito Bezarius <raito@lix.systems>
We were catching ForeignExceptions believing it came from the TRY_AWAIT
handler, but this was misguided.
`j.dump()` is evaluated in synchronous context, outside of the `try {
... } catch (...)` block from `TRY_AWAIT`.
Therefore, we need to use `JSON::Exception` directly.
The previous test case did not catch it because:
(1) https://git.lix.systems/lix-project/lix/issues/865 hid the fact that
`--arg` was wrong.
(2) we did not grep for the warning because… we were not even copying
the strange store path to the binary cache.
(3) checking for the NAR happened after the NAR directory was emptied
for test reasons and this was not even caught neither.
Anyway, the test case was completely busted and has now been tested
without this commit and after this commit and we can confirm that prior
to this commit, the test will fail with an exception trace.
Co-authored-by: Maximilian Bosch <maximilian@mbosch.me>
Change-Id: I8df5befd06c4a449072b987f82a67bc4437e7e49
Signed-off-by: Raito Bezarius <raito@lix.systems>
PushActivity does not work with async code since we have no such thing
as promise-local storage. it will be confusing at best, and completely
wrong at worst, with the current thread-local linking state. if we can
find a way to get promise-local storage we may want to bring this back
though, explicit context passing is rather error-prone. luckily we are
not using parent links for anything important, just to keep the multi-
line activity display from filling up with stuff we're already showing
Change-Id: Ie373d713080a3db811b2d5abd681f78137735e45
checking that the remote build actually succeeded only implied-trusted
remotes or CA derivations makes *absolutely* no sense. we should check
that builds have succeeded before trying to copy them from the remote.
Change-Id: Ib2cf216c580f4c577dd9fef8849acc033ae082b9
Including the `.cc` is possible but is suspicious.
Change-Id: Ie18fef1e30e517edff4ab96f4a9c339e2b3145b5
Signed-off-by: Raito Bezarius <raito@lix.systems>
postInstall needs to know the name of the (versioned) .so file since it
needs to be renamed for Darwin platforms. Unfortunately, the .so version
is not properly overrideable, so we need to use string replacement.
Change-Id: Idf9671f84fac955a52d82a20ec0f381d05fdc762
lowdown 1.4.0 changed the lowdown_opts to include a new and separate
lowdown_opts_term which allows for configuring values specific to
-Tterm (which we're using). This version should have been called 2.0.0
according to semver, hence 2.0.0 was released later without any actual
breaking changes to sort of migitate the problem.
We need to support lowdown >= 1.3 && < 1.4 since the ship has sailed for
updating lowdown in NixOS 25.05 as well as lowdown >= 1.4 or we'll be
stuck in Nixpkgs forever. Support for < 1.4 can be dropped as soon as
NixOS 25.05 is EOL, assuming this change lands before NixOS 25.11
branch-off.
We detect the changed API based on the lowdown version from pkg-config
and define LOWDOWN_SEPARATE_TERM_OPTS based on that. The ifdef is named
according to the specific API change that impacts us, so that it's
hopefully a little simpler to maintain going forward. In the new API,
all newly configurable settings use what would have been the (implicit)
default before. Changing some of these values, especially hpadding,
could be interesting in future changes.
Compared to cl/3081, this change makes sure to initialize all new fields
of lowdown_opts_term explicitly.
It seems that, while making -Tterm more configurable, lowdown's word
wrapping behavior changed slightly which broke basic_repl.test. I've
chosen to work around this by using builtins.add as an example which has
a very short documentation string, so wrapping doesn't matter.
Change-Id: Id73be4c0e43d7eb4f56e10a261b4254402698ff8
capnp does not handle fd passing correctly in all circumstances. we hit
such cirumstances when passing large closures path lists to build-hook.
since capnp seems to ignore fds passed in non-final segments of any rpc
message we just ensure that the capability including the log fd will be
small enough to not be fragmented on the receiving side of the channel.
cf https://github.com/capnproto/capnproto/issues/2359
Change-Id: Id22309264936b3a57bcc68a0753c3bfb3c9a43d2
Some users may have arbitrary needs to connect to their store URIs, e.g.
mTLS authentication, Kerberos authentication, custom renewal using any
RPC mechanism of their preference and so on.
To avoid encoding all these patterns in Lix itself, we push the
configuration to the plugin boundaries and offer a hook for end users to
inherit from `HttpBinaryCacheStore` and provide new store schemes like
`https+mtls://my.very.secure.cache?tls-certificate=...&tls-key=...` or
`https+krb5://my.kerberos.enabled.cache`.
Co-authored-by: George Shammas <george@shamm.as>
Co-authored-by: eldritch horrors <pennae@lix.systems>
Signed-off-by: Raito Bezarius <raito@lix.systems>
Change-Id: I79f322b1a74632500fc79d53f5c920f9e43fd0c4
Usually, EOFs are represented by returning 0 in the `read` APIs, at
least, this is what read(2) dictate.
As clever creature, we may sum zeroes sometimes (advanced form:
`buf->added(got)`) and forego handling the EOF condition.
To avoid the bug that lurked in remote-store.cc and caused busy looping
if the remote end disconnects suddenly, we return
`Result<Option<size_t>>` forcing the caller to perform a specific
processing for the EOF situation.
The conversion did not raise any other offending code path.
Change-Id: I185fdcb77aa82d87ab0802d66ac37c1363657a73
Signed-off-by: Raito Bezarius <raito@lix.systems>
Also adds an assert that store path hash part length is what we expect
because it's alarmingly easy to forget to truncate a hash before
throwing it into there. It's kind of messy code, someone could improve
it more later.
Change-Id: I5296ea3d5b854323d092f0256defb598dd5b87e8
remote builds failures used to be signaled via exit status 1 of the
build hook, which in turn only happened because the build errors we
got from remote stores was thrown and bubbled up to main which then
logged the error and exited with code 1. with rpc we cannot do this
any more. barring a rewrite of the worker infra to allow for errors
being reported with something other than process exit codes this is
the best can do. ideally we would wrap remote builds in a new goal.
(and then remove all exit code shenanigans from DerivationGoal too)
fixes#928
Change-Id: Idc3ede3cbaca34c8c8e40247da52794f2a5013b9
drop our reimplementation of splice for non-linux in favor of using kj
pumpTo. this avoids select() for its O(maxfd) behavior, and if kj ever
uses something more efficient than read/write loops we'll benefit too.
Change-Id: Id01ba84bf8831455af2d9755bf1a3039d215bb47
libarchive *should* not break with 0710 on the tmpdir root on darwin,
just like it doesn't break on linux, but for some reason it does. the
restriction to 0710 can be weakened to 0750 with causing any trouble.
fixes#921
Change-Id: Ia9fc2f8eb9695fc19cefae9857368d5a4e58c8b9
previously we only had one build hook in waiting at most because build
hook rpc was synchronous. now that it no longer is we attempt to start
one hook per derivation, which depending on scheduling can be a *very*
large number. restrict the waiting hook count to 4 to some concurrency
without collecting a large number of hooks that may never do anything.
Change-Id: Ic0b1125cec4acd69e8a0d4639c232e71b825e01d
although we only chown if the build was requested by a local daemon
user. daemonless invocations will not chown as they do not have to.
remote builds *can* chown to the remote builder user, but that does
not seem to happen (for some reason keep-failed is not propagated).
Change-Id: Ic0ead406b38b4ca0556fec42d84888efa25123bf
this makes the actual build directories used by builders invisible and
inaccessible to other processes on the system, avoiding another vector
for outside processes to interfere with builds or pass credentials the
build sandbox should not have access to into the build sandbox anyway.
fixes#919
Change-Id: Ifaa4d8e3940cfde1406e925f75c1375d2e86d81a
this touches both libutil and libstore because with no rpc users it
doesn't make that much sense to separate the two. note that all our
strings are represented as Data (ie, blobs) because capnp Text must
be nul-terminated. while it's technically possible to use Text with
strings containing non-terminating NULs it is a bit of a hassle and
could lead to rpc users erroneously stopping at the first NUL byte.
Change-Id: I4c75e03b79a226ffa8d7cd985e3ac632a0cd7c1c
we need this to generate dependency information, and it'll be the entry
point for custom codegen once we need it. a wrapper also makes it a lot
easier to generate a whole namespace's worth of rpc definitions at once
Change-Id: Iba7a1c92a8a40bede9ed71aa3ab455477ff5e568
if the hook accepts the build request we can handle the entire request
in tryBuildHook. there is no need to punt a partially handled build to
the caller (we only did this to minimize churn during asyncification).
Change-Id: Iec3e35a8103da4fc5fbef394cc28a134ee62a198
mapping the result of an await operation before unpacking it lets us
inject rpc type conversion functions without duplicating all that is
needed for proper exception wrapping and async error traces support.
Change-Id: Ibcba1cc6d2b275757e3475881ef20f95dd4d684f
Goals:
- Distribute reviews to people who can do the reviews
- Not prevent anything from getting done
- Allow giving away more commit access
Anti-goals:
- Silo people into particular areas
- Discourage contributing to any area
This was drafted by glancing at git logs. It is not likely to be very
accurate; the goal here is that we figure out a way to distribute
reviews to the right people.
Change-Id: I8be44bf7fdeca23da8099124eec7bc3a30e34627
`Outcome<void, T>` and `Result<std::optional<T>>` can be interpreted as
being the same thing, but the latter is easier to use: not only do they
allow TRY_AWAIT usage for their promises, we also don't have the error/
exception confusion of outcomes (where the T above is the "error" type)
Change-Id: I92c9241481cecc97e2992445b3dced53c82a2524
while this does require spawning a thread for every contended lock now
we don't expect performance to be impacted. only build-remote used the
synchronous method, and it only used it to serialize uploads to remote
builders. these uploads are expensive enough to dwarf the thread cost.
Change-Id: Iad0aa0cd738bc96fd06a90d655803dadffa09c47
DerivationGoal::InputStream existed only because we did not have an
error-reporting AsyncInputStream of our own yet. we do have one now
though and can thus delete old code in favor of the generic variant
Change-Id: I01c7c564554f8794bdf54603b239b7a808faeda0
it's effectively unused. one use is a write and a read immediately after
the write, the other use checks whether it's not equal to itself (..wat)
Change-Id: I5f6ce26e75a6bfa500c2e9ac3fc70e8dafc9bd74
Commits 205c59367c and
325e7e1824 introduced real glibc store paths from
current nixpkgs unstable into the source. Since nixpkgs `fetchFromGitea` (and
similar fixed-output derivations) depends on a C library, on x86_64-linux they
will fail with the forbidden reference error:
error: the fixed-output derivation '/nix/store/wnmnj3jzc82y89sfmyicr04kilg8zs2k-source.drv' must not reference store paths but 1 such references were found:
/nix/store/q4wq65gl3r8fy746v9bbwgx4gzn0r2kl-glibc-2.40-66
Falsify the store path to prevent this failure.
Change-Id: I949033567bcad070f9a0a19cefdb33a79222e421
My lix build failed today with this result:
```
lix> [----------] 3 tests from MonitorFdHup
lix> [ RUN ] MonitorFdHup.works
lix> [ OK ] MonitorFdHup.works (0 ms)
lix> [ RUN ] MonitorFdHup.works_with_pipes
lix> stderr:
lix> Using configuration: seed=6402097764877502971
lix> libc++abi: terminating due to uncaught exception of type std::__1::future_error: The state of the promise has already been set.
lix> 4/5 lix:check / libstore-unit-tests OK 1.10s
lix> 5/5 lix:check / libexpr-unit-tests OK 1.12s
lix> Summary of Failures:
lix> 3/5 lix:check / libutil-unit-tests FAIL 0.93s killed by signal 6 SIGABRT
lix> Ok: 4
lix> Fail: 1
lix> Full log written to /nix/var/nix/builds/nix-build-lix-2.94.0-dev-pre20250711-65ef28d.drv-0/source/build/meson-logs/testlog.txt
```
I had a response best described as "wtf". I think the cause of this
problem is that there's a race condition with the test in which the loop
gets gone around again a second time because it's triggered by the
terminate fd (and I guess the flags remained what they were before?
seems reasonable), and this is probably racing with the quit atomic
being first to break out of the loop.
I don't know how many hundreds of lixes I've compiled without my test
failing, but this is definitely a bug. I don't think this affects actual
usage as the only impact is repeat delivery of Ctrl-C which is harmless
and which users do regularly.
Change-Id: I60da81d4ac2e79052cd323b5171f9d8bd0aa6783
this was a mess. ssh:// remotes used the extra static fds for build
logs, ssh-ng:// remotes did not. ssh-ng remotes did not use them at
all since ssh-ng never redirected them to begin with. we now create
pipes dynamically and only for ssh:// builders, then translate logs
received over these pipes into the same format used by ssh-ng. this
requires a new activity we did not have before, but since we have a
great many activities that rarely show up already this shouldn't be
a problem for external tooling. if anything external tools can tell
what's going on much better now (at least for ssh:// remote builds)
Change-Id: I02010cee45598362a947faa3a5b04800d39daa31
Avoids incorrect behavior with large integers in `elemAt`, `substring`,
`genList`, etc, which results into crashing the Lix interpreter.
At the same time, unit tests were added for these edge cases with 2^32
as an argument of these primops.
Port of https://github.com/NixOS/nix/pull/13309.
Prior art in https://github.com/NixOS/nix/pull/7222 (forgotten by the original project…).
Change-Id: I1c43ed64f26bcb60e51869e11a74e5de2b7db53a
Co-authored-by: Raito Bezarius <raito@lix.systems>
Signed-off-by: Raito Bezarius <raito@lix.systems>
async queries easily lead to high contention on the localstore sqlite
lock. optimizing the lock wakeup scheme improves query performance by
a linear factor (with the O(waiters) wakeup replaced by O(1) wakeup).
on 100k drv closures we're now at 55s query, down from >8min in 2.93.
Change-Id: I9b96e792c4518a782c690dea92e61260f08f0bad
That way it's easier to spot whether a node is the "final" node in the
graph which is especially helpful for larger graphs.
Change-Id: I460a699f07f5455917792599f4247ebf8f430d93
Closes#334Closes#626
This is loosely based on upstream PR#10877[1], but heavily changed to
use the graph logic from `nix why-depends`.
`precise` is `false` here since the out-path of the drv being built
isn't registered yet, so the path accessor cannot scan through files
yet.
Example output (from an openssh build with `pcsclite.lib` & `glibc` in
`disallowedRequisites`):
error: output '/nix/store/hr8lmmjmd1jk6s3p5ymggyk4am7n2lmb-openssh-10.0p2' is not allowed to refer to the following paths:
/nix/store/p6r5awz3ywrz66symnrn0xb85xzmcysf-pcsclite-2.3.0-lib
/nix/store/q4wq65gl3r8fy746v9bbwgx4gzn0r2kl-glibc-2.40-66
Shown below are chains that lead to the forbidden path(s).
/nix/store/hr8lmmjmd1jk6s3p5ymggyk4am7n2lmb-openssh-10.0p2
└───/nix/store/ys91ywnwikm14xznwk3cdbprapv2m37z-libfido2-1.16.0
└───/nix/store/p6r5awz3ywrz66symnrn0xb85xzmcysf-pcsclite-2.3.0-lib
/nix/store/hr8lmmjmd1jk6s3p5ymggyk4am7n2lmb-openssh-10.0p2
├───/nix/store/q4wq65gl3r8fy746v9bbwgx4gzn0r2kl-glibc-2.40-66
├───/nix/store/6r4zqb04fq5l5l4zghq76wvcpz7dwd35-linux-pam-1.6.1
│ ├───/nix/store/q4wq65gl3r8fy746v9bbwgx4gzn0r2kl-glibc-2.40-66
[...]
[1] https://github.com/NixOS/nix/pull/10877
Co-authored-by: Robert Hensing <robert@roberthensing.nl>
Change-Id: Ib30024c0d9e45c1160bf0134f7d3ba17dbdeff47
While working on the LocalDerivationGoal code, I realized that this
attribute is only set to `false`/`true` depending on whether
`__structuredAttrs` is `true`/`false`.
Change-Id: I53868cd32cedd7e25cb6233bd93bc01111b56a07
This will be useful for other things as well such as the
disallowedRequisites error in the builder code. Additionally, print the
dependencyPath in the tree bold to spot where a change terminates.
Also implemented some unit-tests for this code.
Change-Id: I8460f3f6c5095d5bfbe390f223bc0252800dca5e
The Node struct should become an implementation detail when moving this
into libstore. A map from a node to its direct references is more
intuitive here.
Change-Id: I9fddce6b398b8bb97834e5586bee72b244885fdd
* Better name for refs
* Use std::optional<T> for distance
Suggested-by: eldritch horrors <pennae@lix.systems>
Change-Id: Ie35c3f2a7ea1a90ce3a9807025d0af9ea73e2403
Instead of logging directly, we now write into a `Strings` set that is
referenced by the caller.
While at it, added a test-case to ensure that self-reference invocations
and --all behave properly.
Change-Id: Ib183ab8e8e90436300e1c870fb3ae8f18730abbf
That way we get a line of output per test completed,
which makes it more obvious it's actually doing things.
Change-Id: Ifbbe8bdf64e7178d3c59349cf071eb5a9d0fcd32
staging-next banned !structuredAttrs && separateDebugInfo && disallowedRequisites
due to weird output interactions. Enable structuredAttrs so we can build again.
Also, fix type confusion that makes stdenv explode (https://github.com/NixOS/nixpkgs/issues/422989).
Co-authored-by: eldritch horrors <pennae@lix.systems>
Change-Id: Ic0c773394ee79e10d427f27750d59892d6d1f1d1
this partially reverts commit 0cc021ee15,
which for some reason is completely broken on darwin: there seems to be
no way to receive process-directed signals on a non-main thread. trying
to do it anyway will fail silently. since we only ever used kj for this
to get signal handling timeouts on darwin (which lacks sigtimedwait) to
print a nice message about retrying ^C again we can work around this by
moving the message printing into a fresh, unrelated, non-signal thread.
Change-Id: I5939c6ec62a7e1dc1b3f16067f77277533949fa0
We missed xokdvium being author on cl/3300. This is something we
absolutely want to avoid.
We credit xokdvium in the RL and add a note on this problem.
Thanks to xokdvium for reaching out in private to us so we can repair
this mistake.
Change-Id: I094d0f95b6647104621d6b228e69a4529a300304
Signed-off-by: Raito Bezarius <raito@lix.systems>
many a cleanup path has been broken by interruptions being thrown every
time checkInterrupt is called. we should only throw *once* though; more
than one Interrupted exception for the same event is not only confusing
but also breaks all cleanup paths at the first checkInterrupt call site
(e.g. #900, the cgroup cleanup saga, temp dirs not being removed, etc).
Change-Id: Ibfabf7f6af6ac2b78ad93582c254bbc48fcb3073
we must be crash-safe *anyway*, and being unable to interrupt lix if it
gets stuck somewhere that never calls checkInterrupt is really annoying
Change-Id: I7c40271c3da7e69d8735e22b7b7c4751b5306ab6
macos doesn't have sigtimedwait and we need signal wait timeouts in
order to print a "please hit ^C again" message with a bit of delay.
Change-Id: If574fb1a9de0b19975b34fc63662b089eaedc9d2
another checkInterrupt can be a makeInterruptible wrapper now. this is
also necessary to add a second daemon socket for the new rpc protocol.
Change-Id: I55055f975335a75708f1f73edb75f7bfe77a5938
2025-07-04 21:12:12 +02:00
1213 changed files with 24814 additions and 13514 deletions
Following the changes in 2.94.0 to shorten build directory paths, aimed at [resolving UNIX domain socket length issues](https://gerrit.lix.systems/c/lix/+/4168/13) and [improving nix-shell](https://git.lix.systems/lix-project/lix/issues/940), we inadvertently introduced an excessively long path for the `$NIX_BUILD_TOP` environment variable used by Nix shells (their effective temporary `/build` directory).
To fix this, we replaced the `build-top-$HASH` directory name with simply `build-top`, reducing these paths by at least 30 characters.
We also added a test to ensure that Nix shells do not introduce more than 50 extra characters relative to their base directory (e.g., `/tmp` when `$TMPDIR` is not set).
synopsis:"Add `builtins.warn` for emitting warnings from Nix code"
cls:[2248]
category:"Features"
credits:[milibopp, Qyriad]
---
Lix now has a builtin function for emitting warnings.
Like `builtins.trace`, it takes two arguments: the message to emit, and the expression to return.
_Unlike_`builtins.trace`, `builtins.warn` requires the first argument — the message — to be a string.
In the future we may extend `builtins.warn` to accept a more structured API.
To go along with this, we also have two new config settings:
- [`debugger-on-warn`](@docroot@/command-ref/conf-file.md#conf-debugger-on-warn), which, when used with `--debugger`, makes `builtins.warn` also function like [`builtins.break`](@docroot@/language/builtins.md#builtins-break).
- [`abort-on-warn`](@docroot@/command-ref/conf-file.md#conf-abort-on-warn), which aborts evaluation entirely after the warning is emitted.
If you need to disable cgroups temporarily, remember that you can do
`NIX_CONF='include /etc/nix/nix.conf\nuse-cgroups = false' nix-build ...` or
`nix-build --no-use-cgroups ...`.
## What about other service managers than systemd?
systemd has a [documentation](https://systemd.io/CGROUP_DELEGATION/) on how to
handle cgroup delegation from service management perspective.
If your service manager adheres to systemd semantics, e.g. writing an extended
attribute `user.delegate=1` on the delegated cgroup tree directory and moving
the `nix-daemon` process inside a cgroup tree to respect the inner process
rule, then, the feature will work as well.
## Why is the cgroup feature still experimental?
While the cgroup feature unlocks many use cases, its behavior and integration (e.g. user experience), especially at scale on build farms or in multi-tenant environments, are not yet fully matured. There’s also potential for deeper systemd integration (e.g. using slices and scopes) that has not been fully explored.
To avoid locking in an unstable interface, we’re keeping the experimental flag until we have validated the feature across a broader range of scenarios, including but not limited to:
synopsis:"Deprecate shadowing internal files through the Nix search path"
issues:[998]
cls:[4632]
category:"Breaking Changes"
credits:[thubrecht]
---
As Lix uses the path `<nix/fetchurl.nix>` for bootstrapping purposes, the ability to shadow it by adding `nix=/some/path` (or `/other/path` that contains a `nix` directory) to the search path is not desirable.
To alleviate potential issues, Lix now emits a warning when the Nix search path contains potential shadows for internal files, which will be changed to an error in a future release.
The warning can be disabled by enabling the deprecated feature `nix-path-shadow`.
1. These builds are not run via the daemon, which owns `/nix/var/nix/builds`.
2. The user lacks permissions for that path.
We considered making `build-dir` a store-level option and defaulting it to `<chroot-root>/nix/var/nix/builds` for chroot stores, but opted instead for a fallback: if the default fails, Nix now creates a safe build directory under `/tmp`.
To avoid CVE-2025-52991, the fallback uses an extra path component between `/tmp` and the build dir.
**Note**: this fallback clutters `/tmp` with build directories that are not cleaned up. To prevent this, explicitly set `build-dir` to a path managed by Lix, even for local workloads.
You're not alone. Thousands of Lix users suffer every day from excessive `builtins.toString` syndrome. It’s 2025, and we still have to cast integers to use them in strings.
To address this, Lix introduces the **`coerce-integers`** experimental feature. When enabled, interpolated integers within `"${...}"` are automatically coerced to strings. This allows writing:
To enable the feature, you need to add `coerce-integers` to your set of experimental features.
### Stabilization criteria
The `coerce-integers` feature is experimental and limited strictly to string interpolation (`"${...}"`). Before stabilization, the following must hold:
1.**Interpolation-only**
Coercion must not occur outside interpolation. Expressions like `"" + 42` must continue to fail.
2.**Expectation that no explicit cast are being observed**
Cases observing explicit coercion (e.g., via `tryEval` gadget or similar) are expected not to be load-bearing in actual production code.
### Timeline for stabilization
If the feature proves safe and is widely adopted across typical usage (e.g., actual configurations in the wild turning on the flag, non-trivial out-of-tree projects using it), the experimental flag will be removed **after six months of active use or two Lix releases**, whichever is longer.
This avoids locking the feature in experimental status indefinitely, as happened with Flakes, while allowing time for validation and ecosystem integration.
### What about coercing floats or more?
Coercion beyond integers -- such as for floats or other types -- is **not planned**, even under an experimental flag. Questions like "what is the canonical string representation of a float?" involve subtle and context-dependent trade-offs. Without a robust and principled mechanism to define and audit such behavior, introducing broader coercion risks setting unintended and hard-to-reverse precedents. The scope of `coerce-integers` is intentionally narrow and will remain so.
In terms of outlook, a proposal like https://git.lix.systems/lix-project/lix/issues/835 could pave the way for a better solution.
synopsis:"Add an indication of nix-shell nesting depth"
cls:[4657]
issues:[fj#826]
category:"Improvements"
credits:[thubrecht]
---
When in a nix shell (either via a `nix-shell` or a `nix develop` invocation), a variable `NIX_SHELL_LEVEL` is exported to indicate the nesting depth of nix shells.
synopsis:Remove reliance on Bash for remote stores via SSH
issues:[fj#830, fj#805, fj#304]
cls:[3159]
category:"Fixes"
credits:[raito]
---
The pre-flight `echo started` handshake -- added years ago to catch race conditions -- has been removed.
After removal of connection sharing in Lix 2.93, it required a Bash-compatible shell and a standard `echo`, so it failed on:
* builders protected by `ForceCommand` wrappers (e.g. `nix-remote-build`),
* BusyBox / initrd images with no Bash,
* hosts using non-POSIX shells such as Nushell.
The race the probe once addressed was tied to SSH connection-sharing -- since connection-sharing code has already been removed, the probe is now pointless.
Real connection or protocol errors are now left to SSH/Nix to report directly.
This is technically a breaking change if you had scripts that relied on the literal "started" which needs to be updated to rely on other signals, e.g., exit codes.
synopsis:"repl-overlays now work in the debugger for flakes"
issues:[fj#777]
cls:[3398]
category:Fixes
credits:[jade]
---
Due to a bug, it was previously not possible to use the debugger on flakes with repl-overlays, or with pure evaluation in general:
```
$ nix repl --pure-eval
Lix 2.94.0-dev-pre20250617-87d99da
Type :? for help.
Loading 'repl-overlays'...
error: access to absolute path '/Users/jade/.config/nix/repl.nix' is forbidden in pure eval mode (use '--impure' to override)
```
This is now fixed.
The contents of the repl-overlays file itself (i.e. most typically the top level lambda in it) will be evaluated in impure mode.
It may be necessary to use `builtins.seq` to force the impure operations to happen first if one wants to do impure operations inside a repl-overlays file in pure evaluation mode.
> If you are building via the Lix daemon (default on Linux and macOS), it is the Lix daemon user account (that is, `root`) that should have SSH access to a user (not necessarily `root`) on the remote machine.
>
> Furthermore, `root` needs to have the public host keys for the remote system in its `.ssh/known_hosts`.
> To add them to `known_hosts` for root, do `ssh-keyscan USER@HOST | sudo tee -a ~root/.ssh/known_hosts`.
> To add them to `known_hosts` for root, do `ssh-keyscan HOST | sudo tee -a ~root/.ssh/known_hosts`.
>
> If you can’t or don’t want to configure `root` to be able to access the remote machine, you can use a private Nix store instead by passing e.g. `--store ~/my-nix` when running a Nix command from the local machine.
## Configuration
The list of remote machines can be specified on the command line or in
the Lix configuration file. The former is convenient for testing. For
example, the following command allows you to build a derivation for
- `uri`: Due to backward compatibility, the `ssh://` may be omitted for the store-uri.
- `ssh-public-host-key`: The key must be provided encoded in base64. Specifically calculated via `base64 -w0 /etc/ssh/ssh_host_ed25519_key.pub`
### Format detection
At first, the given configuration is being parsed syntactically as a toml.
If parsing fails and the given configuration contains a `"` the error is presented to the user, as those characters are necessary for TOML, but disallowed for the legacy format.
Otherwise, parsing is retried using the legacy format.
If non-syntactic errors are detected within the toml, the exception will always be shown to the user directly.
## Builder selection
The configuration(s) above specify several machines that can perform `i686-linux` builds.
However, `granite` will only do builds that have the attribute
```nix
requiredSystemFeatures = [ "benchmark" ];
```
or
```nix
requiredSystemFeatures = [ "benchmark" "kvm" ];
```
`diorite` cannot do builds that require `kvm`, but `andesite` does support
such builds. For regular builds, `diorite` will be preferred over
# [Pasta](https://passt.top/passt/about/): a network sandbox for fixed-output derivations
## Introduction
This section only applies to **Linux systems** as Pasta is a Linux-only measure.
Since [CVE-2025-46416](https://lix.systems/blog/2025-06-24-lix-cves/), the Lix project decided to adopt [Pasta](https://passt.top/passt/about/) for all fixed-output derivations, protecting against various attack vectors such as UNIX abstract domain sockets or more manipulation at the network layer from a malicious fixed-output derivation code.
Pasta acts as a translation layer between a layer-2 network interface and layer-4 sockets (TCP, UDP, ICMP/ICMPv6 echo) on the host. It requires no special privileges and can serve as a alternative to [SLiRP](https://en.wikipedia.org/wiki/Slirp) which was used [by Guix to mitigate the same problem](https://codeberg.org/guix/guix/commit/fb42611b8f27960304db5a1c0d33b8371dcde2a8).
## How to disable Pasta?
It's sufficient to pass `pasta-path = ""` in your `/etc/nix/nix.conf` or on the command line `--pasta-path ""` of a Lix invocation.
## Known issues surrounding Pasta
- Only the first DNS server in `/etc/resolv.conf` is considered: failover is not possible.
- [Reduced feature set compared to the Linux kernel](https://passt.top/passt/about/#features)
- [Performance overhead in multi-gigabits contexts and IMIX MTUs](https://passt.top/passt/about/#performance_1)
@@ -58,7 +58,7 @@ $ nix-build flake:nixpkgs -A firefox
$ nix-build flake:github:NixOS/nixpkgs/release-23.11 -A firefox
```
Finally, for legacy reasons, if a path starts with `channel:`, the rest of the argument is interpreted as the name of a *nixpkgs* channel tarball to fetch from `https://nixos.org/channels/$CHANNEL_NAME/nixexprs.tar.xz`.
Finally, for legacy reasons, if a path starts with `channel:`, the rest of the argument is interpreted as the name of a *nixpkgs* channel tarball to fetch from `https://channels.nixos.org/$CHANNEL_NAME/nixexprs.tar.xz`.
This is a **hard coded URL** pattern and is *not* related to the subscribed channels managed by the [nix-channel](./nix-channel.md) command.
> **Note**: any of the special syntaxes may always be disambiguated by prefixing the path.
Channels are a mechanism for referencing remote Nix expressions and conveniently retrieving their latest version.
The moving parts of channels are:
- The official channels listed at <https://nixos.org/channels>
- The official channels listed at <https://channels.nixos.org>
- The user-specific list of [subscribed channels](#subscribed-channels)
- The [downloaded channel contents](#channels)
- The [Nix expression search path](@docroot@/command-ref/conf-file.md#conf-nix-path), set with the [`-I` option](#opt-I) or the [`NIX_PATH` environment variable](#env-NIX_PATH)
@@ -77,9 +77,9 @@ This command has the following operations:
Subscribe to the Nixpkgs channel and run `hello` from the GNU Hello package:
A typical development flow for simple changes in Lix looks like:
- [Set up and build Lix](#building)
- For large changes, check in regarding design and possibly create an RFD issue on Forgejo
- Make the changes in your editor
- [Send the changes to Gerrit](#sending-to-gerrit)
- Once you have the number for the CL from Gerrit to put in the changelog, [write a changelog entry](#release-notes) and amend it into the commit
- Update the Gerrit change by submitting it with the same command as the first time
- Request and receive a code review
- Address feedback from the review
- Amend commits, send to Gerrit again
- Submit the approved change
## Building Lix in a development shell {#building}
### Setting up the development shell
@@ -48,7 +60,7 @@ $ just setup build test-unit
$ just install test-integration
```
Many targets have a `-custom` variant which pass extra arguments to `meson`.
Many justfile aliases have a `-custom` variant which pass extra arguments to `meson`.
For example, to work on both Lix and nix-eval-jobs you can run:
```
@@ -129,7 +141,59 @@ To inspect the canonical source of truth on what the state of the buildsystem co
$ meson introspect
```
## Building Lix outside of development shells
#### LLD
The development shell on Linux uses LLD by default for faster link times.
This is set using `mesonFlags`, so to override it, you can simplify re-specify the linker to Meson:
```bash
$ just setup-custom -Dc_link_args=-fuse-ld=ld -Dcpp_link_args=-fuse-ld=ld
```
While using LLD, you may find it helpful to use ThinLTO for even further improvements to link times for incremental builds:
```bash
$ just setup-custom -Db_lto=true -Db_lto_mode=thin -Db_thinlto_cache=true
```
## Sending changes to Gerrit for review {#sending-to-gerrit}
We use Gerrit for all our code review in Lix.
Our instance is at <https://gerrit.lix.systems>.
There's much more information about how to use Gerrit in the [wiki section on Gerrit][wiki-gerrit] including how to use Jujutsu, how to use the UI and more.
The Snix project also has some Gerrit information [in their contributing docs][snix-gerrit].
Then, you can request a review via the "Reply" button on the web UI.
If you click "Suggest Owners", it will try to suggest the maintainers of the area of the code change to send review requests to.
Requesting reviews from multiple people is normal.
We do our best to respond to directly sent reviews in a few days, so feel free to request another reviewer or ask on Matrix if you've not got a response for a while.
Keep in mind that Lix is a volunteer project and we have limited bandwidth, so some changes aren't feasible to shepherd through; please check in on Matrix at design time when doing large changes.
Once you get a `Code-Review+2` vote on your change, it's rebased on `main` and CI marks it `Verified+1`, you're able (and usually expected, so you can have a second chance to check it over) to hit the Submit button to merge it.
If the change appears as "Rebase Required", you need to rebase it on `main` locally or via the Gerrit UI and wait for `Verified+1` before the Submit button is made active
The `Code-Review+2` from before will stick around through trivial rebases so no need to re-request review for a mere rebase.
## Interacting with the CI, Buildkite
We use Buildkite for our CI, usually you will not have to interact directly with it other than reviewing any errors it produces, which are linked from Gerrit.
However in certain cases a CI run will fail due to transient issues not related to your code and you will need to rerun it by hand.
You can log in to the CI via [SSO](https://buildkite.com/sso/lix-project). On your job you can then hit the "Retry failed" button to rerun it, normally you will not have a repeat of the transient issue.
If the build still fails on CI issues or all builds are failing this should be reported via [Zulip on #T-infra](https://zulip.lix.systems/#narrow/channel/7-T-infra) or [Matrix on #dev](https://matrix.to/#/%23dev%3Alix.systems?via=lix.systems).
## Building Lix with `nix`
To build a release version of Lix for the current operating system and CPU architecture:
@@ -286,10 +350,10 @@ Configure your editor to use the `clangd` from the shell, either by running it i
> Some other editors (e.g. Emacs, Vim) need a plugin to support LSP servers in general (e.g. [lsp-mode](https://github.com/emacs-lsp/lsp-mode) for Emacs and [vim-lsp](https://github.com/prabirshrestha/vim-lsp) for vim).
> Editor-specific setup is typically opinionated, so we will not cover it here in more detail.
### Checking links in the manual
# Manual and documentation
## Building the manual
The build checks for broken internal links.
This happens late in the process, so `nix build` is not suitable for iterating.
The built manual is in `build/doc/manual/manual/index.html`.
`@docroot@` provides a base path for links that occur in reusable snippets or other documentation that doesn't have a base path of its own.
The build checks for broken internal links.
This happens late in the process, so `nix build` is not suitable for iterating and it's recommended to use the `meson` command above instead.
If a broken link occurs in a snippet that was inserted into multiple generated files in different directories, use `@docroot@` to reference the `doc/manual/src` directory.
### `@\docroot\@` variable
If the `@docroot@`literal appears in an error message from the `mdbook-linkcheck` tool, the `@docroot@` replacement needs to be applied to the generated source file that mentions it.
See existing `@docroot@` logic in the [Makefile].
Regular markdown files used for the manual have a base path of their own and they can use relative paths instead of`@docroot@`.
`@\docroot\@`provides a base path for links that occur in reusable snippets or other documentation that doesn't have a base path of its own.
If a broken link occurs in a snippet that was inserted into multiple generated files in different directories, use`@\docroot\@` to reference the `doc/manual/src` directory.
If the `@\docroot\@` literal appears in an error message from the `mdbook-linkcheck` tool, the `@\docroot\@` replacement needs to be applied to the generated source file that mentions it.
See existing `@\docroot\@` logic in `doc/manual/substitute.py`.
Regular markdown files used for the manual have a base path of their own and they can use relative paths instead of `@\docroot\@`.
## API documentation
@@ -341,7 +410,7 @@ You can build it yourself:
Metrics about the change in line/function coverage over time will be available in the future (FIXME(lix-hydra)).
## Add a release note
## Add a release note {#release-notes}
`doc/manual/rl-next` contains release notes entries for all unreleased changes.
@@ -410,15 +479,15 @@ The following properties are supported:
### Build process
Releases have a precomputed `rl-MAJOR.MINOR.md`, and no `rl-next.md`.
Set `buildUnreleasedNotes = true;` in `flake.nix` to build the release notes on the fly.
Development releases have a generated `rl-next.md`.
## Adding experimental or deprecated features, global settings, or builtins
# Adding experimental or deprecated features, global settings, or builtins
Experimental and deprecated features, global settings, and builtins are generally referenced both in the code and in the documentation.
To prevent duplication or divergence, they are defined in data files, and a script generates the necessary glue.
The data file format is similar to the release notes: it consists of a YAML metadata header, followed by the documentation in Markdown format.
### Experimental or deprecated features
## Experimental or deprecated features
Experimental and deprecated features support the following metadata properties:
*`name` (required): user-facing name of the feature, to be used in `nix.conf` options and on the command line.
@@ -428,7 +497,7 @@ Experimental and deprecated features support the following metadata properties:
Experimental feature data files should live in `lix/libutil/experimental-features`, and deprecated features in `lix/libutil/deprecated-features`.
They must be listed in the `experimental_feature_definitions` or `deprecated_feature_definitions` lists in `lix/libutil/meson.build` respectively to be considered by the build system.
### Global settings
## Global settings
Global settings support the following metadata properties:
*`name` (required): user-facing name of the setting, to be used as key in `nix.conf` and in the `--option` command line argument.
@@ -456,7 +525,7 @@ Settings are not collected in a single place in the source tree, so an appropria
Look for related setting definition files under second-level subdirectories of `lix` whose name includes `settings`.
Then add the new file there, and don't forget to register it in the appropriate `meson.build` file.
### Builtin functions
## Builtin functions
The following metadata properties are supported for builtin functions:
*`name` (required): the language-facing name (as a member of the `builtins` attribute set) of the function.
@@ -472,7 +541,7 @@ The following metadata properties are supported for builtin functions:
New builtin function definition files must be added to `lix/libexpr/builtins` and registered in the `builtin_definitions` list in `lix/libexpr/meson.build`.
### Builtin constants
## Builtin constants
The following metadata properties are supported for builtin constants:
*`name` (required): the language-facing name (as a member of the `builtins` attribute set) of the constant.
*`type` (required): the Nix language type of the constant; the C++ type is automatically derived.
@@ -62,6 +62,12 @@ For `installcheck` specifically, first run `just install` before running the tes
Finer-grained filtering within a test suite is also possible using the [--gtest_filter](https://google.github.io/googletest/advanced.html#running-a-subset-of-the-tests) command-line option to a test suite executable, or the `GTEST_FILTER` environment variable.
### Inspecting failures
The test suite emits logs in `build/meson-logs/`; the full textual failure logs are in `build/meson-logs/testlog.txt`.
If you want a much nicer experience of viewing the logs in a structured manner, use `xunit-viewer --results build/meson-logs/testlog.junit.xml --server` to view them in a web browser.
### Unit test support libraries
There are headers and code which are not just used to test the library in question, but also downstream libraries.
@@ -395,7 +401,6 @@ I grepped `lix/` for `get[eE]nv\("` to find the mentions in Lix code.
**Expected value**: the path to an executable shell
- `PRINT_PATH` - Undocumented. Used by `nix-prefetch-url` as an alternative form of `--print-path`. Why???
- `_NIX_IN_TEST` - If present with any value, makes `fetchClosure` accept file URLs in addition to HTTP ones. Why is this not `_NIX_FORCE_HTTP`??
Not used anywhere else.
- `NIX_ALLOW_EVAL` - Used by eval-cache tests to block evaluation if set to `0`.
Lix in Docker is very sensitive to **functional** DNS resolution if you are running with [Pasta protections](../advanced-topics/pasta.md) which are enabled by default since Lix 2.93.0 on most distributions.
If you notice failure to download things, double check whether your **first** DNS entry in `/etc/resolv.conf` is functional.
Lix with [Pasta protections](../advanced-topics/pasta.md) does not support failing over the next entries.
In Lix, only builds done on Linux with `sandbox = true` and a functioning
`pasta-path` are isolated from the rest of the system, all other builds are
considered non-isolated to some degree.
For example, running Lix with [Pasta](@docroot@/advanced-topics/pasta.md)
disabled makes the host network visible to fixed-output derivations, reducing
isolation somewhat.
## Clean termination of non-isolated builds
Non-isolated builds may not terminate cleanly in all cases due to limitations in Lix's process management.
This occurs when a build keeps the build log file descriptor open past the end of the actual build. A common cause of this are background tasks that aren't properly terminated before the main build process exits, for example: HTTP servers run as part of a test suite.
See [issue #1018](https://git.lix.systems/lix-project/lix/issues/1018) for an example.
The only solution is to manually terminate leftover processes in your derivation, including during failure scenarios.
FIXME(Lix): This section does not document the most common modern practices in terms of avoiding channels, pinning, declarative software installation (see flakey-profile or home-manager or NixOS), or using flakes, etc.
It is, however, likely correct at a technical level.
For more information on modern practices, see the [resources](https://lix.systems/resources) page on the Lix site.
For more information on modern practices, see the [resources](https://wiki.lix.systems/books/lix-users/page/nix-resources) page on the Lix site.
</div>
@@ -41,7 +41,7 @@ install Lix. If this is not the case for some reason, you can add it
- Remove support for daemon protocols before 2.18 [fj#510](https://git.lix.systems/lix-project/lix/issues/510) [cl/3249](https://gerrit.lix.systems/c/lix/+/3249)
Support for daemon wire protocols belonging to Nix 2.17 or older have been
removed. This impacts clients connecting to the local daemon socket or any
remote builder configured using the `ssh-ng` protocol. Builders configured
with the `ssh` protocol are still accessible from clients such as Nix 2.3.
Additionally Lix will not be able to connect to an old daemon locally, and
remote build connections to old daemons is likewise limited to `ssh` urls.
We have decided to take this step because the old protocols are very badly
tested (if at all), maintenance overhead is high, and a number of problems
with their design makes it infeasible to remain backwards compatible while
we move Lix to a more modern RPC mechanism with better versioning support.
Many thanks to [eldritch horrors](https://git.lix.systems/pennae) for this.
- Remove impure derivations and dynamic derivations [fj#815](https://git.lix.systems/lix-project/lix/issues/815) [cl/3210](https://gerrit.lix.systems/c/lix/+/3210)
The `impure-derivations` and `dynamic-derivations` experimental feature have
been removed.
New impure or dynamic derivations cannot be created from this point forward, and
any such pre-existing store derivations canot be read or built any more.
Derivation outputs created by building such a derivation are still valid
until garbage collected; existing store derivations can only be garbage
collected.
Many thanks to [eldritch horrors](https://git.lix.systems/pennae) for this.
- First argument to `--arg`/`--argstr` must be a valid Nix identifier [fj#496](https://git.lix.systems/lix-project/lix/issues/496)
The first argument to `--arg`/`--argstr` must be a valid Nix identifier, i.e.
`nix-build --arg config.allowUnfree true` is now rejected.
This is because that invocation is a false friend since it doesn't set
The idea is to change the behavior to the latter in the long-term. For that,
non-identifiers started giving a warning since 2.92 and are now rejected to give people
who depend on that a chance to notice and potentially weigh in on the discussion.
Many thanks to [ma27](https://git.lix.systems/ma27) for this.
- New cgroup delegation model [fj#537](https://git.lix.systems/lix-project/lix/issues/537) [fj#77](https://git.lix.systems/lix-project/lix/issues/77) [cl/3230](https://gerrit.lix.systems/c/lix/+/3230)
Builds using cgroups (i.e. `use-cgroups = true` and the experimental feature
`cgroups`) now always delegate a cgroup tree to the sandbox.
Compared to the original C++ Nix project, our delegation includes the
`subtree_control` file as well, which means that the sandbox can disable
certain controllers in its own cgroup tree.
This is a breaking change because this requires the Nix daemon to run with an
already delegated cgroup tree by the service manager.
## How to setup the cgroup tree with systemd?
systemd offers knobs to perform the required setup using:
```
[Service]
Delegate=yes
DelegateSubtree=supervisor
```
These directives are now included in our systemd packaging.
## What about using Nix as root without connecting to the daemon?
Builds run as `root` without connecting to the daemon relying on the cgroup
feature are now broken, i.e.
```console
# nix-build --use-cgroups --sandbox ... # will not work
If you need to disable cgroups temporarily, remember that you can do
`NIX_CONF='include /etc/nix/nix.conf\nuse-cgroups = false' nix-build ...` or
`nix-build --no-use-cgroups ...`.
## What about other service managers than systemd?
systemd has a [documentation](https://systemd.io/CGROUP_DELEGATION/) on how to
handle cgroup delegation from service management perspective.
If your service manager adheres to systemd semantics, e.g. writing an extended
attribute `user.delegate=1` on the delegated cgroup tree directory and moving
the `nix-daemon` process inside a cgroup tree to respect the inner process
rule, then, the feature will work as well.
## Why is the cgroup feature still experimental?
While the cgroup feature unlocks many use cases, its behavior and integration (e.g. user experience), especially at scale on build farms or in multi-tenant environments, are not yet fully matured. There’s also potential for deeper systemd integration (e.g. using slices and scopes) that has not been fully explored.
To avoid locking in an unstable interface, we’re keeping the experimental flag until we have validated the feature across a broader range of scenarios, including but not limited to:
* Nix as root
* Hydra-style build farms
* Forgejo CI runners
* Shared remote builders
Many thanks to [Raito Bezarius](https://git.lix.systems/raito), [eldritch horrors](https://git.lix.systems/pennae), and [lheckemann](https://git.lix.systems/lheckemann) for this.
- Enable high compress ratio zstd compression by default for binary caches uploads [fj#945](https://git.lix.systems/lix-project/lix/issues/945) [cl/4503](https://gerrit.lix.systems/c/lix/+/4503)
The default compression method for binary cache uploads has been switched from
[`xz`](https://github.com/tukaani-project/xz) to
[`zstd`](https://github.com/facebook/zstd) to address performance and usability
issues related to modern hardware and high-speed connections.
## Why?
`xz` offers compression ratios but is single-threaded in our implementation and
very slow (~10-20 Mbps in our test), preventing full utilization of 100Mbps+
connections and significantly slowing decompression for end users.
Lix is a "compress once, decompress many" application: build farms can afford
to spend more time compressing to achieve a faster download transfer for the
end user. More importantly, it matters that all end users spend the least
amount of time decompressing.
## What about compression ratios?
`zstd` cannot achieve the same peaks as `xz`, nonetheless, `zstd` compression
level has been increased to level 12 by default to balance compression ratio
and performance.
## Synthetic test case data
* **xz** (default compression level) on a 4.4GB file: ~632MB (77s)
* **zstd** (level 12) on the same file: ~775MB (18s), 18% larger but 50% faster
* **zstd** (level 14): ~773MB (37s)
* **zstd** (level 16): ~735MB (66s)
Many thanks to [eldritch horrors](https://git.lix.systems/pennae) and [Raito Bezarius](https://git.lix.systems/raito) for this.
- Repl debugger uses `--ignore-try` by default [lix#666](https://git.lix.systems/lix-project/lix/issues/666) [cl/3488](https://gerrit.lix.systems/c/lix/+/3488)
Previously, using the debugger meant that exceptions thrown in `builtins.tryEval` would trigger the debugger.
However, this caught nixpkgs initialization code, which is unhelpful in the majority of cases, so we changed the default.
We consider this a breaking change since eval results *will* change if strings
with embedded NUL bytes were used, but we also consider the old behavior to be
not intentional (seeing how inconsistent it was) but merely fallout from a old
and misguided implementation decision to be worked around, not actually fixed.
Many thanks to [eldritch horrors](https://git.lix.systems/pennae) for this.
- Fixed output derivations can be run using `pasta` network isolation [fj#285](https://git.lix.systems/lix-project/lix/issues/285) [cl/3452](https://gerrit.lix.systems/c/lix/+/3452)
Fixed output derivations traditionally run in the host network namespace.
On Linux this allows such derivations to communicate with other sandboxes
or the host using the abstract Unix domains socket namespace; this hasn't
been unproblematic in the past and has been used in two distinct exploits
to break out of the sandbox. For this reason fixed output derivations can
now run in a network namespace (provided by [`pasta`]), restricted to TCP
and UDP communication with the rest of the world. When enabled this could
be a breaking change and we classify it as such, even though we don't yet
enable or require such isolation by default. We may enforce this in later
releases of Lix once we have sufficient confidence that breakage is rare.
[`pasta`]: https://passt.top/
Many thanks to [eldritch horrors](https://git.lix.systems/pennae) and [puck](https://git.lix.systems/puck) for this.
- Function equality semantics are more consistent, but still bad [cl/4556](https://gerrit.lix.systems/c/lix/+/4556) [cl/4244](https://gerrit.lix.systems/c/lix/+/4244)
Lix has inherited a historic misfeature from CppNix in the form of pointer
equality checks built into the `==` operator. These checks were originally
meant to optimize comparison for large sets, but they have the unfortunate
side effect of producing unexpected results when sets containing functions
are compared. **Lix 2.93 and earlier** behave as shown in the repl session
```
Lix 2.93.3
Type :? for help.
nix-repl> f = x: x
Added f.
nix-repl> f == f
false
nix-repl> let s.f = f; in s.f == s.f
false
nix-repl> # however!
{ inherit f; } == { inherit f; }
true
nix-repl> [ f ] == [ f ]
true
nix-repl> # and, in another twist:
[ f ] == map f [ f ]
false
```
Nixpkgs relies on sets containing functions being comparable, so we cannot
simply deprecate this behavior. Due to changes to the object model used by
Lix ***all* comparisons above now evaluate to `true`**. This is considered
a breaking change because eval results may differ, but we also consider it
minor because the optimization is unsound (c.f. `let l = [NaN]; in l == l`
evaluates to `true` even though floating point `NaN` is incomparable). Lix
intends to remove this optimization altogether in the future, but until we
can do that we instead make it slightly less broken to allow other, *real*
optimizations. Function equality comparison remains **undefined behavior**
and should not be relied upon in Nixlang code that intends to be portable.
Many thanks to [eldritch horrors](https://git.lix.systems/pennae) for this.
- `nix eval --write-to` has been removed [fj#974](https://git.lix.systems/lix-project/lix/issues/974) [fj#227](https://git.lix.systems/lix-project/lix/issues/227) [cl/4045](https://gerrit.lix.systems/c/lix/+/4045)
`nix eval --write-to` has been removed since it was underspecified, not widely
useful, and prone to security-sensitive misbehaviors. The feature was added in
Nix 2.4 purely for internal use in the build system. According to our research
it hasn't found any use outside of some distribution packaging scripts. Please
use structured outputs formats (such as JSON) instead as they have better type
fidelity, don't conflate attributes with paths, and are useful to other tools.
Many thanks to [eldritch horrors](https://git.lix.systems/pennae) for this.
- Remove the `parse-toml-timestamps` experimental feature
The `parse-toml-timestamps` experimental feature has been removed.
This feature used in‐band signalling to mark timestamps, making it
impossible to unambiguously parse TOML documents. It also exposed
implementation‐defined behaviour in the TOML specification that
changed in the toml11 parser library.
Any interface for parsing TOML timestamps suitable for future
stabilization would necessarily involve breaking changes, and there
is no evidence this experimental feature is being relied upon in the
wild, so it has been removed.
Many thanks to [Emily](https://git.lix.systems/emilazy) for this.
* Add `http3 = true` in your Nix configuration for permanent activation.
To disable it, use `--no-http3`.
**Note**:
* `--no-http2 --http3` will still enable both HTTP/2 and HTTP/3.
* `--http2 --http3` will prioritize HTTP/3 and fall back to HTTP/2 (and then
HTTP/1.1).
These are current CLI limitations. In the future, we plan to replace `--httpX`
options with `--max-http-version [1,2,3]` for easier version selection in Lix
transfers.
Many thanks to [Raito Bezarius](https://git.lix.systems/raito) and [eldritch horrors](https://git.lix.systems/pennae) for this.
- Add hyperlinks in attr set printing [cl/3790](https://gerrit.lix.systems/c/lix/+/3790)
The attribute set printer, such as is seen in `nix repl` or in type errors, now prints hyperlinks on each attribute name to its definition site if it is known.
Example: all of the attributes shown here are hyperlinks to the exact definition site of the attribute in question:
You're not alone. Thousands of Lix users suffer every day from excessive `builtins.toString` syndrome. It’s 2025, and we still have to cast integers to use them in strings.
To address this, Lix introduces the **`coerce-integers`** experimental feature. When enabled, interpolated integers within `"${...}"` are automatically coerced to strings. This allows writing:
To enable the feature, you need to add `coerce-integers` to your set of experimental features.
### Stabilization criteria
The `coerce-integers` feature is experimental and limited strictly to string interpolation (`"${...}"`). Before stabilization, the following must hold:
1. **Interpolation-only**
Coercion must not occur outside interpolation. Expressions like `"" + 42` must continue to fail.
2. **Expectation that no explicit cast are being observed**
Cases observing explicit coercion (e.g., via `tryEval` gadget or similar) are expected not to be load-bearing in actual production code.
### Timeline for stabilization
If the feature proves safe and is widely adopted across typical usage (e.g., actual configurations in the wild turning on the flag, non-trivial out-of-tree projects using it), the experimental flag will be removed **after six months of active use or two Lix releases**, whichever is longer.
This avoids locking the feature in experimental status indefinitely, as happened with Flakes, while allowing time for validation and ecosystem integration.
### What about coercing floats or more?
Coercion beyond integers -- such as for floats or other types -- is **not planned**, even under an experimental flag. Questions like "what is the canonical string representation of a float?" involve subtle and context-dependent trade-offs. Without a robust and principled mechanism to define and audit such behavior, introducing broader coercion risks setting unintended and hard-to-reverse precedents. The scope of `coerce-integers` is intentionally narrow and will remain so.
In terms of outlook, a proposal like https://git.lix.systems/lix-project/lix/issues/835 could pave the way for a better solution.
Many thanks to [Raito Bezarius](https://git.lix.systems/raito), [delroth](https://github.com/delroth), [eldritch horrors](https://git.lix.systems/pennae), and [winter](https://git.lix.systems/winter) for this.
- nix-eval-jobs: support `--no-instantiate` flag [fj#987](https://git.lix.systems/lix-project/lix/issues/987)
`nix-eval-jobs` now supports a flag called `--no-instantiate`. With this enabled,
no write operations on the eval store are performed. That means, only evaluation is
performed, but derivations (and their gcroots) aren't created.
Many thanks to [mic92](https://github.com/mic92) and [ma27](https://git.lix.systems/ma27) for this.
## Improvements
- Assess current profile generations pointers in `nix doctor` [cl/3108](https://gerrit.lix.systems/c/lix/+/3108)
Added a new check to `nix doctor` that verifies whether the current generation of
a Nix profile can be resolved. This helps users diagnose issues with broken or
misconfigured profile symlinks.
This helps determining if you have broken symlinks or misconfigured packaging.
Many thanks to [Raito Bezarius](https://git.lix.systems/raito) for this.
- Improved susbtituter query speed
The code used to query substituters for derivations has been rewritten slightly
to take advantage of our asynchronous runtime. Such queries run for every build
that could download from substituters and processes every derivation that isn't
yet present on the local system. Previously Lix would use `http-connections` to
limit query concurrency, even for modern caches that support HTTP/2 and have no
limit on how many queries can be run concurrently on one single connection. Lix
no longer does this, resulting in approximately 60% reduction in query time for
medium-sized closures (e.g. NixOS system closures) during testing, although the
exact number depends greatly on local network latency and generally improves as
latency increases. Unlike previously setting `http-connections` to `1` or other
low values no longer brings a massive penalty in query performance if the cache
in use by the querying system supports HTTP/2 (as e.g. `cache.nixos.org` does).
Many thanks to [eldritch horrors](https://git.lix.systems/pennae) for this.
Hitting Control-C or sending `SIGINT` to Lix now prints an informational message
if it is still running after on second, the second Control-C/`SIGINT` terminates
Lix immediately without waiting for any shutdown code to finish running. Lix did
not treat the second such event differently from first in the past; this made it
impossible to easily terminate running Lix processes that got stuck in e.g. very
expensive Nixlang code that never interacted with the store. We now terminate as
soon as the user hits Control-C again without waiting any more, to much the same
effect as putting Lix into the background and killing it immediately afterwards.
This means you can now more conveniently break out of stuck Nixlang evaluations:
```
❯ nix-instantiate --eval --expr 'let f = n: if n == 0 then 0 else f (n - 1) + f (n - 1); in f 32'
^CStill shutting down. Press ^C again to abort all operations immediately.
^C
❌130 ❯
```
Many thanks to [eldritch horrors](https://git.lix.systems/pennae) for this.
- `--keep-failed` chowns the build directory to the user that request the build
Running a build with `--keep-failed` now chowns the temporary directory from the
builder user and group to the user that request the build if the build came from
a local user connected to the daemon. This makes inspecting failed derivations a
lot easier. On Linux the build directory made visible to the user will not be in
the same path as it was in the sandbox and continuing builds will usually break.
Many thanks to [eldritch horrors](https://git.lix.systems/pennae) for this.
- Better debuggability on fixed-output hash mismatches
Fixed-output derivation hash mismatch error messages will now include the path that was
produced unexpectedly, and this path will be registered as valid even if `--check`
(`nix-store`, `nix-build`) or `--rebuild` (`nix build`) was passed. This makes comparing
the expected path with the obtained path easier, and is useful for debugging when
upstreams modify previously-published releases or when changes in fixed-output
derivations' dependencies affect their output unexpectedly.
Many thanks to [lheckemann](https://git.lix.systems/lheckemann) for this.
- Add --raw flag to `nix-instantiate --eval` for unescaped output [gh#12119](https://github.com/NixOS/nix/pull/12119) [cl/2886](https://gerrit.lix.systems/c/lix/+/2886)
The `nix-instantiate --eval` command now supports a `--raw` flag. When used,
the result must be coercible to a string (as with `${...}`) and is printed
verbatim, without quotes or escaping.
Many thanks to [Martin Fischer](https://github.com/not-my-profile), [infinisil](https://github.com/infinisil), and [Raito Bezarius](https://git.lix.systems/raito) for this.
- Allow `nix store ls` to read nar listings from binary cache stores. [cl/3225](https://gerrit.lix.systems/c/lix/+/3225)
The `nix store ls` command now supports reading `.ls` nar listings from binary cache stores.
If a listing is detected for the store path being queried, the nar is no longer downloaded.
These nar listings are available in binary cache stores where the `write-nar-listing` option is
enabled, such as cache.nixos.org.
Many thanks to [Victor Fuentes](https://git.lix.systems/vlinkz) for this.
- show tree with references that lead to an output cycle [fj#551](https://git.lix.systems/lix-project/lix/issues/551)
When Lix determines a cyclic dependency between several outputs of a derivation,
it now displays which files in which outputs lead to an output cycle:
```
error: cycle detected in build of '/nix/store/gc5h2whz3rylpf34n99nswvqgkjkigmy-demo.drv' in the references of output 'bar' from output 'foo'.
Shown below are the files inside the outputs leading to the cycle:
Please note that showing the files and its contents while displaying the cycles only works
on Linux.
Many thanks to [ma27](https://git.lix.systems/ma27) for this.
- Lix now enables parallel marking in boehm-gc [fj#983](https://git.lix.systems/lix-project/lix/issues/983) [cl/3880](https://gerrit.lix.systems/c/lix/+/3880)
This brings a fairly modest performance improvement (~38% for `nixpkgs search hello`) to evaluation, especially in scenarios that necessitate larger heap sizes.
Many thanks to [Eelco Dolstra](https://github.com/edolstra) and [Seth Flynn](https://git.lix.systems/getchoo) for this.
- `disallowedRequisites` now reports chains of disallowed requisites [fj#334](https://git.lix.systems/lix-project/lix/issues/334) [fj#626](https://git.lix.systems/lix-project/lix/issues/626) [gh#10877](https://github.com/NixOS/nix/issues/10877)
When a build fails because of [`disallowedRequisites`](@docroot@/language/advanced-attributes.md#adv-attr-disallowedRequisites), the error message now includes the chain of references that led to the failure. This makes it easier to see in which derivations the chain can be broken, to resolve the problem.
Example:
```
$ nix-build -A hello
error: output '/nix/store/0b7k85gg5r28gb54px9nq7iv5986mns9-hello-2.12.2' is not allowed to refer to the following paths:
Many thanks to [ma27](https://git.lix.systems/ma27) and [Robert Hensing](https://github.com/roberth) for this.
- Stack traces now summarize involved derivations at the bottom [cl/4493](https://gerrit.lix.systems/c/lix/+/4493)
When evaluation errors and a stack trace is printed,
For example, if I add Nheko to a NixOS `environment.systemPackages` without adding `olm-3.2.16` `nixpkgs.config.permittedInsecurePackages`, then without `--show-trace`, I previously got this:
```
error:
… while calling the 'head' builtin
at /nix/store/9v6qa656sq3xc58vkxslqy646p0ajj61-source/lib/attrsets.nix:1701:13:
1700| if length values == 1 || pred here (elemAt values 1) (head values) then
1701| head values
| ^
1702| else
… while evaluating the attribute 'value'
at /nix/store/9v6qa656sq3xc58vkxslqy646p0ajj61-source/lib/modules.nix:1118:7:
1117| // {
1118| value = addErrorContext "while evaluating the option `${showOption loc}':" value;
| ^
1119| inherit (res.defsFinal') highestPrio;
(stack trace truncated; use '--show-trace' to show the full trace)
error: Package ‘olm-3.2.16’ in /nix/store/9v6qa656sq3xc58vkxslqy646p0ajj61-source/pkgs/by-name/ol/olm/package.nix:37 is marked as insecure, refusing to evaluate.
< -snip the whole explanation about olm's CVEs- >
```
This doesn't tell me anything about where `olm-3.2.16` came from.
With `--show-trace`, there's 1155 lines to sift through, but does contain lines like "while evaluating derivation 'nheko-0.12.1'".
With this change, those lines are summarized and collected at the bottom, regardless of `--show-trace`:
```
error:
… while calling the 'head' builtin
at /nix/store/9v6qa656sq3xc58vkxslqy646p0ajj61-source/lib/attrsets.nix:1701:13:
1700| if length values == 1 || pred here (elemAt values 1) (head values) then
1701| head values
| ^
1702| else
… while evaluating the attribute 'value'
at /nix/store/9v6qa656sq3xc58vkxslqy646p0ajj61-source/lib/modules.nix:1118:7:
1117| // {
1118| value = addErrorContext "while evaluating the option `${showOption loc}':" value;
| ^
1119| inherit (res.defsFinal') highestPrio;
(stack trace truncated; use '--show-trace' to show the full trace)
error: Package ‘olm-3.2.16’ in /nix/store/9v6qa656sq3xc58vkxslqy646p0ajj61-source/pkgs/by-name/ol/olm/package.nix:37 is marked as insecure, refusing to evaluate.
< -snip the whole explanation about olm's CVEs- >
note: trace involved the following derivations:
derivation 'etc'
derivation 'dbus-1'
derivation 'system-path'
derivation 'nheko-0.12.1'
derivation 'mtxclient-0.10.1'
```
Now we finally know that olm was evaluated because of Nheko, without sifting through *thousands* of lines of error message.
Many thanks to [Qyriad](https://git.lix.systems/Qyriad) for this.
- Symbols reuses once-allocated Value to reduce garbage collected allocations [cl/3308](https://gerrit.lix.systems/c/lix/+/3308) [cl/3300](https://gerrit.lix.systems/c/lix/+/3300) [cl/3314](https://gerrit.lix.systems/c/lix/+/3314) [cl/3310](https://gerrit.lix.systems/c/lix/+/3310) [cl/3312](https://gerrit.lix.systems/c/lix/+/3312) [cl/3313](https://gerrit.lix.systems/c/lix/+/3313)
In the Lix evaluator, **symbols** represent immutable strings, like those used
for attribute names.
In evaluator design, such strings are typically [**interned**](https://en.wikipedia.org/wiki/String_interning), stored uniquely
to save memory, and Lix inherits this approach from the original C++ codebase.
However, some builtins, like `builtins.attrNames`, must return a `Value` type
that can represent any Nix value (strings, integers, lists, etc.).
Before this change, these builtins would create lists of `Value` objects by
allocating them through the garbage collector, copying the symbol’s string
content each time.
This allocation is unnecessary if the interned symbols themselves also hold a
`Value` representation allocated outside the garbage collector, since these
live for the full duration of evaluation.
As a result, this reduces the number of allocations, leading to:
* A significant drop in maximum [resident set memory](https://en.wikipedia.org/wiki/Resident_set_size) (RSS), with some large-scale
tests showing up to 11% (about 500 MiB) savings in large colmena deployments.
* A slight decrease in CPU usage during Nix evaluations.
This change is inspired by https://github.com/NixOS/nix/pull/13258 but the approach is different.
**Note** : [`xokdvium`](https://github.com/xokdvium) is the rightful author of https://gerrit.lix.systems/c/lix/+/3300 and the credit was missed on our end during the development process. We are deeply sorry for this mistake.
Many thanks to [Raito Bezarius](https://git.lix.systems/raito), [eldritch horrors](https://git.lix.systems/pennae), [Tom Hubrecht](https://git.lix.systems/tom-hubrecht), [xokdvium](https://github.com/xokdvium), and [NaN-git](https://github.com/NaN-git) for this.
## Fixes
- `build-dir` no longer defaults to `temp-dir` [cl/3453](https://gerrit.lix.systems/c/lix/+/3453)
The directory in which temporary build directories are created no longer defaults
to the value of the `temp-dir` setting to avoid builders making their directories
world-accessible. This behavior has been used to escape the build sandbox and can
cause build impurities even when not used maliciously. We now default to `builds`
in `NIX_STATE_DIR` (which is `/nix/var/nix/b` in the default configuration).
Many thanks to [eldritch horrors](https://git.lix.systems/pennae) for this.
- Global certificate authorities are copied inside the builder's environment [gh#12698](https://github.com/NixOS/nix/issues/12698) [fj#885](https://git.lix.systems/lix-project/lix/issues/885) [cl/3765](https://gerrit.lix.systems/c/lix/+/3765)
Previously, CA certificates were only installed at
`/etc/ssl/certs/ca-certificates.crt` for sandboxed builds on Linux.
This setup was insufficient in light of recent changes in `nixpkgs`, which now
enforce HTTPS usage for `fetchurl`, even for fixed-output derivations, to
mitigate confidentiality risks such as `netrc` or credentials leakage.
`nixpkgs` still make use of a special package called `cacerts` which contains a
copy of the CA certificates maintained by Nixpkgs and added as a reference for
TLS-enabled fetchers.
As a result, having a consistent and trusted certificate authority in all
builder environments is becoming more essential.
On `nix-darwin`, the `NIX_SSL_CERT_FILE` environment variable is always
explicitly defined, but it is ignored by the sandbox setup.
Simultaneously, Nix evaluates and propagates impure environment variables via
`lib.proxyImpureEnvVars`, meaning that if `NIX_SSL_CERT_FILE` is set (which
influences the default value for `ssl-cert-file`), it will be forwarded
unchanged into the builder environment.
However, on Linux, Nix also *copies* the CA file into the sandbox, creating a
discrepancy between the value of `NIX_SSL_CERT_FILE` and the actual trusted
certificate path used during the build.
This divergence caused confusion and was partially addressed by attempts to
whitelist the CA path in the Darwin sandbox (see cl/2906), but that approach
involved a non-trivial path canonicalization step and is not as general as this one.
To address this properly, we now emit a warning and override
`NIX_SSL_CERT_FILE` inside the builder, explicitly pointing it to the CA file
copied into the sandbox.
This eliminates ambiguity between `NIX_SSL_CERT_FILE`
and `ssl-cert-file`, ensuring consistent trust anchors across platforms.
This warning might become a hard error as we figure out what to do regarding
`lib.proxyImpureEnvVars` in nixpkgs.
The behavior has been verified across sandboxed and unsandboxed builds on both
Linux and Darwin.
As a consequence of this change, approximately 500KB of CA certificate data is
now unconditionally copied into the build directory for fixed-output
derivations.
While this ensures consistent trust verification without having to restart the
daemon after system upgrades, it may introduce a slight overhead in build
performance. At present, no optimizations have been implemented to avoid this
copy, but if this overhead proves noticeable in your workflows, please open an
issue so we can evaluate and possibly implement different strategies to render
trust anchors visible.
Many thanks to [Raito Bezarius](https://git.lix.systems/raito) and [Emily](https://git.lix.systems/emilazy) for this.
- libstore: exponential backoff for downloads [lix#932](https://git.lix.systems/lix-project/lix/issues/932) [cl/3856](https://gerrit.lix.systems/c/lix/+/3856)
The connection timeout when downloading from e.g. a binary cache is exponentially
increased per failure. The option `connect-timeout` is now an alias to `max-connect-timeout`
which is the maximum value for a timeout. The start value is controlled
by `initial-connect-timeout` which is `5` by default.
Many thanks to [ma27](https://git.lix.systems/ma27) for this.
- Fix develop shells for derivations with escape codes [fj#991](https://git.lix.systems/lix-project/lix/issues/991) [cl/4154](https://gerrit.lix.systems/c/lix/+/4154) [cl/4155](https://gerrit.lix.systems/c/lix/+/4155)
ASCII control characters (including `\e`, used for ANSI escape codes) in derivation variables are now correctly escaped for `nix develop` and `nix print-dev-env`, instead of erroring.
Many thanks to [Qyriad](https://git.lix.systems/Qyriad) for this.
To find out why, use nix-store --query --roots and nix-store --query --referrers.
```
Many thanks to [lheckemann](https://git.lix.systems/lheckemann) for this.
- Fallback to safe temp dir when build-dir is unwritable [fj#876](https://git.lix.systems/lix-project/lix/issues/876) [cl/3501](https://gerrit.lix.systems/c/lix/+/3501)
Non-daemon builds started failing with a permission error after introducing the `build-dir` option:
1. These builds are not run via the daemon, which owns `/nix/var/nix/builds`.
2. The user lacks permissions for that path.
We considered making `build-dir` a store-level option and defaulting it to `<chroot-root>/nix/var/nix/builds` for chroot stores, but opted instead for a fallback: if the default fails, Nix now creates a safe build directory under `/tmp`.
To avoid CVE-2025-52991, the fallback uses an extra path component between `/tmp` and the build dir.
**Note**: this fallback clutters `/tmp` with build directories that are not cleaned up. To prevent this, explicitly set `build-dir` to a path managed by Lix, even for local workloads.
Many thanks to [Raito Bezarius](https://git.lix.systems/raito) and [eldritch horrors](https://git.lix.systems/pennae) for this.
- Parse overflowing JSON number literals as floating‐point [cl/3919](https://gerrit.lix.systems/c/lix/+/3919)
Previously, `builtins.fromJSON "-9223372036854775809"` would
return a floating‐point number, while `builtins.fromJSON
"9223372036854775808"` would cause an evaluation error. This was
introduced with the banning of integer overflow in Lix 2.91; previously
the latter would result in C++ undefined behaviour. These cases are
now treated consistently with JSON’s model of a single numeric type,
and JSON number literals that do not fit in a Nix‐language integer
will be parsed as floating‐point numbers.
Many thanks to [Emily](https://git.lix.systems/emilazy) for this.
- Fix handling of OSC codes in terminal output [fj#160](https://git.lix.systems/lix-project/lix/issues/160) [cl/3143](https://gerrit.lix.systems/c/lix/+/3143)
OSC codes in terminal output are now handled correctly, where OSC 8 (hyperlink) is preserved any
time color codes are allowed and all other OSC codes are stripped out. This applies not only to
output from build commands but also to rendered documentation in the REPL.
Many thanks to [lilyball](https://git.lix.systems/lilyball) for this.
- Fix nix develop for derivations that rejects dependencies with structured attrs [fj#997](https://git.lix.systems/lix-project/lix/issues/997) [cl/4182](https://gerrit.lix.systems/c/lix/+/4182)
For the sake of concision, we refer to `disallowedReferences` in what follows,
but all output checks were equally fixed:
`{dis,}allowed{References,Requisites}`.
Derivations can define *output checks* to reject unwanted dependencies, such as
interpreters like `bash` or compilers like `gcc`. This can be done in two ways:
* **Legacy style**: `disallowedReferences = [ ... ]` in the environment.
`nix-eval-jobs` doesn't clear the `NIX_PATH` from the environment anymore. This matches the behavior
of [upstream version `2.30`](https://github.com/nix-community/nix-eval-jobs/releases/tag/v2.30.0).
Many thanks to [ma27](https://git.lix.systems/ma27) and [mic92](https://github.com/mic92) for this.
- Remove reliance on Bash for remote stores via SSH [fj#830](https://git.lix.systems/lix-project/lix/issues/830) [fj#805](https://git.lix.systems/lix-project/lix/issues/805) [fj#304](https://git.lix.systems/lix-project/lix/issues/304) [cl/3159](https://gerrit.lix.systems/c/lix/+/3159)
The pre-flight `echo started` handshake -- added years ago to catch race conditions -- has been removed.
After removal of connection sharing in Lix 2.93, it required a Bash-compatible shell and a standard `echo`, so it failed on:
* builders protected by `ForceCommand` wrappers (e.g. `nix-remote-build`),
* BusyBox / initrd images with no Bash,
* hosts using non-POSIX shells such as Nushell.
The race the probe once addressed was tied to SSH connection-sharing -- since connection-sharing code has already been removed, the probe is now pointless.
Real connection or protocol errors are now left to SSH/Nix to report directly.
This is technically a breaking change if you had scripts that relied on the literal "started" which needs to be updated to rely on other signals, e.g., exit codes.
Many thanks to [Raito Bezarius](https://git.lix.systems/raito) for this.
- repl-overlays now work in the debugger for flakes [fj#777](https://git.lix.systems/lix-project/lix/issues/777) [cl/3398](https://gerrit.lix.systems/c/lix/+/3398)
Due to a bug, it was previously not possible to use the debugger on flakes with repl-overlays, or with pure evaluation in general:
```
$ nix repl --pure-eval
Lix 2.94.0-dev-pre20250617-87d99da
Type :? for help.
Loading 'repl-overlays'...
error: access to absolute path '/Users/jade/.config/nix/repl.nix' is forbidden in pure eval mode (use '--impure' to override)
```
This is now fixed.
The contents of the repl-overlays file itself (i.e. most typically the top level lambda in it) will be evaluated in impure mode.
It may be necessary to use `builtins.seq` to force the impure operations to happen first if one wants to do impure operations inside a repl-overlays file in pure evaluation mode.
Many thanks to [jade](https://git.lix.systems/jade) for this.
- `nix-shell` default shell directory is not `/tmp` anymore for `$NIX_BUILD_TOP` [fj#940](https://git.lix.systems/lix-project/lix/issues/940)
Previously, Lix `nix-shell`s could exit non-zero status when `stdenv`'s `dumpVars` phase failed to write to `$NIX_BUILD_TOP/env-vars`, despite `dumpVars` being intended as a debugging aid.
This happens when `TMPDIR` is not set and defaults therefore to `/tmp`, resulting in a `/tmp/env-vars` global file that every `nix-shell` wants to write.
We fix this issue by reusing a pre-created, unique, and writable location, as the build top directory, avoiding shell exiting from write failures silently.
Many thanks to [Raito Bezarius](https://git.lix.systems/raito) for this.
When using e.g. [Snix's nar-bridge](https://snix.dev/docs/components/overview/#nar-bridge) via
an `http`-store, Lix would create cache entries with a wrong URL to the NAR when uploading
a store-path.
This caused hard build failures for Hydra.
Lix doesn't create these entries on upload anymore. Instead, it only removes negative cache entries.
The cache entry for a narinfo is now created the first time, Lix queries the cache
for the previously uploaded store-path again.
Many thanks to [ma27](https://git.lix.systems/ma27) for this.
- Lix libraries can now be linked statically [fj#789](https://git.lix.systems/lix-project/lix/issues/789) [cl/3775](https://gerrit.lix.systems/c/lix/+/3775) [cl/3778](https://gerrit.lix.systems/c/lix/+/3778)
Previously the pkg-config files distributed with Lix were only suitable for dynamic linkage, causing "undefined reference to…" linker errors when trying to link statically.
Private dependency information has now been added to make static linkage work as expected without user intervention.
In addition, relevant static libraries are now prelinked to avoid strange failures due to missing static initializers.
Many thanks to [alois31](https://git.lix.systems/alois31) for this.
- add description to zsh completions [fj#910](https://git.lix.systems/lix-project/lix/issues/910) [cl/3632](https://gerrit.lix.systems/c/lix/+/3632)
Emit descriptions when completing args in zsh completions. This uses the descriptions we already
provided in NIX\_GET\_COMPLETIONS.
Many thanks to [matthewbauer](https://github.com/matthewbauer) for this.
## Miscellany
- Deprecation of CA derivations, dynamic derivations, and impure derivations [fj#815](https://git.lix.systems/lix-project/lix/issues/815)
Content-addressed derivations are now deprecated and slated for removal in Lix 2.94.
We're doing this because the CA derivation system has been a known cause of problems
and inconsistencies, is unmaintained, habitually makes improving the store code very
difficult (or blocks such improvements outright), and is beset by a number of design
flaws that in our opinion cannot be fixed without a full reimplementation from zero.
Dynamic derivations and impure derivations are built on the CA derivation framework,
and owing to this they too are deprecated and slated for removal in another release.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.