the redirections the tests did were once not possible with runProgram2,
but they have been for a while now. we should use them instead of fork.
Change-Id: Ia422c8941ed04a9403dee68bfe938d68952253da
cancelling a get() promise could leave the pool with allegedly in use
resource handles that do not actually exist. this would cause asserts
in ~Pool to fire (and crash the entire program due to bad accounting)
probably fixes#1041
Change-Id: Ic61ab9b658f02f71e8be7577b8a1115600846f86
Fixes#838.
xattrs were historically disabled in 2017, since then, NFS v4.2 gained
support for xattrs and almost all filesystems support xattrs. In
addition, moving the build directory to a xattrs-supporting filesystem
is always an option. Programs may exploit xattrs (including ACLs) during
their build process and testing phase, to better serve these programs,
we will remove this limitation.
Lix will use POSIX ACLs for its UNIX domain sockets in the future and be
one of these programs which will run tests making use of xattrs, while
not writing any xattrs in the derivation outputs themselves.
xattrs are still scrubbed from derivation outputs so it is not possible
to obtain an advantage by writing a security-related xattr to a well
chosen file.
Tests were added to test these scenarios on Linux. Darwin is carefully
excluded because of #1008 and #1090, solving that is welcome. In the
meantime, they are marked xfail.
Change-Id: Ia3255eeb8442e83db4f10dcb5a51cbc368a2550d
Signed-off-by: Raito Bezarius <raito@lix.systems>
one would expect the lang tests to be selectable (using `-k`) using
`eval_okay` `eval_fail` etc though this was not the case so far.
This commit renames the functions to reflect the expected names
Change-Id: I4cd340c093d774feeb4d96820d28e49c665b159f
during the value rewrite we accidentally broke extension of incomplete
primop application. this only shows up when binding on incomplete call
to a primop to a name, binding an incomplete call to *that* to another
name, and then finally calling the second binding with enough args for
a complete primop application. since this only shows up when calling a
primop with three or more args it took a while to surface. we have few
builtins that match this: foldl', replaceStrings, and substring. these
are not used incompletely in this manner very often, so this lingered.
fixes#1102
Change-Id: I218dffc14ae876efc86a86c7eb6c895e2405201c
When --check is ran with no output, this is seen as having *SOME*
invalid outputs, actually, ALL of them are invalid here.
Instead of reporting a confusing error, let's consider ALL of outputs
being invalid to be no output at all available and advise the user to
rerun the build without --check.
If there's *some* (>0) invalid outputs, let's print them exactly to give
a chance to the user to remediate this.
Fixes#485.
Change-Id: I00955ef9ea4f129e2c98d68c73b1e981f90278a0
Signed-off-by: Raito Bezarius <raito@lix.systems>
Yeah uh, I don't think `$shell` is ever set in the functional1
environment. Pretty sure this has been expanding to `''` THIS WHOLE
TIME.
Also `SHELL` is `/sbin/nologin` in that builder. idk why so I'm just
passing `shell` unconditionally in functional's `mkDerivation` helper.
Maybe next we should put `set -euo pipefail` into all these test
builders?
Change-Id: Ie5fb85305932c03a901ba84ffcce48f16a6a6964
macos-builder02 in particular has been having problems for a while now
that no one could reproduce. We believe we have finally found the cause:
missing shebangs, not just in the tests themselves but *also* in the
inline scripts `nix-shell` itself writes.
Fixes#1042.
I believe this will also fix#1093.
See also: https://github.com/NixOS/nix/pull/14778
Change-Id: I54d04a770b8e78484815db88a8bc88776a6a6964
`build_env` is assumed to be a certain trivial environment, e.g. a
singleton search path.
An environment in general is either empty, contain one or more search
paths separated by a colon.
It seems that the intent was to set the `build_shell` first as a
prepended path in the managed environment then extend via the
`build_env` parts.
This fixes a usecase when the `BUILD_TEST_ENV` is non-trivial.
Change-Id: If5b8ab976d867a57ce0b8d29255f64695e30a8b2
Signed-off-by: Raito Bezarius <raito@lix.systems>
This adds a way to control the deprecated features that need to be
enable while evaluating the specific nixpkgs version
Change-Id: Ia2b9a29f1966d4d5f32239a36ba8270330007da9
This avoids rebuilding Lix each time we need to modify the lib tests
(as flake.nix stays unchanged), and adds a way to control the deprecated
features that need to be enable while evaluating the nixpkgs lib
testsuite
Change-Id: I2baa1cf90effcaf9d53337c6a7a5b24d6297178c
Certain evaluation requires to pass impurity arguments, add `flags`
to add more arguments to the instantiation.
Change-Id: I538ca2ff12c07cc66e845b42f81baa87f11b8f0f
Signed-off-by: Raito Bezarius <raito@lix.systems>
Is there a situation where `home` can be None? This hinders the ability
to use it quickly without type gating in the test code itself.
Change-Id: Idd739ddad67e10d6f7a6aad283c0e3b7cd08106a
Signed-off-by: Raito Bezarius <raito@lix.systems>
The prior use of the UID is uninteresting in general for consumers who
wants to learn things about _what_ the build is, not what is the UID of
the build user running it.
Achieving this with a place where a mapping between cgroup names and
build requests are maintained is non trivial because there's no "global"
daemon that keeps track of this information and such a daemon would
require a persistent location to track this.
Rather than solving that, we will just encode the derivation hash inside
the cgroup name. Userspace can simply pull this apart and query Nix to
obtain the original corresponding derivation for more information.
This provides userspace with a way to map cgroup names to pnames for
example. Going further, if you have a map between attribute paths and
derivation hashes, you can power a database of cgroup metrics per
attribute paths by combining all of that.
In addition to this rename, we adopt a systemd-ish convention for
template units (`@`).
As a result, userspace tooling that will scrape cgroup metrics will
discover a consistent path as long as they filter out the UID part,
especially in context of Id6c458aad30eaa08c3609ac8280a7dde8e8f3cf9
change.
This problem is tracked under #1073.
Change-Id: I238d0568a3e4b1ff3057781c0639528d666b4d37
Signed-off-by: Raito Bezarius <raito@lix.systems>
nlohmann has customization points for exception throws. we can use these
instead and wrap json exceptions at the source instead of playing a game
of whack-a-mole with json errors all over the tree. since nlohmann needs
macros set to achieve this we can no longer precompile its headers *and*
must forbid including it anywhere without the proper defines or ordering
of include directives will break lix with ODR violations, if we see them
at all before we get another json-related bug report. a new lint will be
in charge of ensuring this doesn't happen. we also re-allow direct calls
to nlohmann json parsing since error handling is no longer a problem. we
will keep the wrapper for more convenient error context handling though.
fixes#1092
Change-Id: I54ecc14f5bec5e2177729b41c3703216e76cc6a3
It's 2026 and the command handler was entirely manual!
We introduce a basic command handler framework for the REPL that handles
all the previous (implicit) features reasonably well.
This also builds up the infrastructure for nicer features.
Most important feature is that now we can let external plugins adds
commands to the REPL.
There's slight REPL changes:
- errors for debugger commands are more explicit.
- short-hand have a long-hand version.
- help shows command sorted by std::map natural key sorting (based on aliases)
while keeping the special syntaxes for the REPL language at the top.
- sections were introduced for Flakes commands.
- :st was merged into one command with an optional argument marker.
Change-Id: Ibd24f994491bf5d45d4b218fd84c0944ae5050fd
Signed-off-by: Raito Bezarius <raito@lix.systems>
Status quo: We have `force$Type` and `eval$Type`, both which first
produce a value and then do a type check. The type checking logic is not
consistently implemented, with lots of code duplication.
This change does:
- Introduce new `check*` functions which unify the logic (the error
handling unfortunately still needs some duplication for now)
- Make both `force*` and `eval*` use the `check*` function for the
actual type checking
- Inline and dismantle the `eval*` functions for being of little use and
little used. This makes the `ExprOp*::eval` implementations for binary
logic operators more verbose, but IMO that's a good thing: The
implementation now needs to be a lot more explicit about the
short-circuiting semantics, something which was previously hidden behind
the short-circuiting semantics of the C++ language, in a way that could
easily be overlooked and lead to confusion, which is something that
happened to me twice in a year.
- Changes `forceAttrs` and `forceList` to include the context in case
`forceValue` fails (compared to only when the type check fails). This
was done for code consistency, because I could not find any reason why
list and attrs had different semantics here than int, float and bool. So
far the visible change is minimal (see the diff on the err.exp), however
this needs vetting for potential performance regressions.
Change-Id: I33e5c706d46850c9e1126293ee01dab85ba07587
That struct has been annoying me since I've encountered it. There is no
reason for having to track all these symbols out of band in the state.
Change-Id: I5c2c0d8174af0a51f9b456cc3651d8203a406d09
I was pretty confused by this when I saw it, not realizing that `build`
meant `tests/functional2/build` and not the top-level Meson `build`
directory.
Change-Id: I4865f196e5bf029419c6b488e78cbfb46a6a6964
...and add both a static bash and busybox into the sandbox for building:
$ /nix/store/fz43jxs5qfg5vldzk38y2hmrcl58qk18-busybox-1.36.1/bin/sh -c 'declare -A foo'
/nix/store/fz43jxs5qfg5vldzk38y2hmrcl58qk18-busybox-1.36.1/bin/sh: declare: not found
The reason is that busybox's `sh` doesn't know about `declare` which
means it cannot build derivations with `__structuredAttrs = true;`.
Change-Id: Ie3ca431f862fd3e59dc649582704ae739ee5834a
Nearly each tests gets a fresh store (in the test root residing under TMPDIR),
which with default settings means 8 MiB of reserved space to be deleted by the
user in case of space shortages, which of course for a test suite is just
wasted. Stop doing that to considerably reduce the litter (experimentally, from
3.7 GiB to 67 MiB per run). The old functional test suite already did the same.
Change-Id: I912cdda6e796ea37cfee0b8fe0478e976a6a6964
this is only needed on macos, and only for tests that build anything. on
linux it can actually *break* stuff due to unfortunate interactions with
store url parsing, which is required to work for any remote build tests.
Change-Id: Ic6f7e090f15e129fc365d7edc56cdbc1a5686047
apparently this file slipped through the code review (the actual test
does exist, no worries) but this empty file and init py exist, sitting
here empty. This commit drops them as they serve no purpose
see commit 1b7ad3a7ad
Change-Id: I64c4e03cd3ceedf9e430ae014127966bd8e4269d
- test_check used an old version of an error message
- test_good2 did not build an unspecified dependency
- test_attribute_selection used a fixed drv path. we
can't keep this with the current setup because the
drv `system` makes the drv hash platform-dependent
fixes#1081
Co-Authored-By: kloenk <me@kloenk.dev>
Change-Id: Ia009740f8b5432e83467ea451a5ff53d5e141b3f
Commentator2.0