notably the bad-store test previously tested *bash*, not lix. we also
need to extend the list of non-fatal errors, but that's probably fine
Change-Id: I983f4adc8047fe800323099de4037711869ee547
this should improve error messages when connecting to long socket paths,
and make it at all possible to detect ENOTDIR from connect helpers. this
will be needed to handle daemon sockets that live in directories we have
not previously considered like the protocol name directories we need for
rpc purposes, all of which would have resulted in hard errors previously
Change-Id: I4e16cf09fedd07c93cfdf7b0f151106984507315
- launch-builder-linux.cc was missing an include for musl
and used function that are not defined in the launchers
- musl caches pids used for raise, breaking sandbox setup
- the mtls contrib plugin won't build, didn't try fixing,
static builds can't really use plugins reliably anyway.
Change-Id: I5ab1664e45ea977e5bcf05e41d825e6014e62146
kj can and will throw exceptions immediately without wrapping them as
promise results, and we must take care of these as well to not crash.
fixes#1150
Change-Id: Ifa29d0f3da4a6490f3e45d4c04a8fa49eaa20f43
not doing this has lix crash if stdio forwarding fails for odd reasons,
such as ssh failing to set up a connection due to transient dns errors.
cf #1148
Change-Id: I543b7444236e69d3c13b47fb6c6eb37c65b96763
`nix.full_path()` is the *built* path, not the *installed* path. in the
sandbox this point to a location that contains *only* `nix`, not any of
the old-cli symlinks. any test that needs them in path would thus fail.
Change-Id: Ia651db8d1151a84f8d1b52e8993f6db61b0fe104
we already report errors that happen during reads via exceptions
rethrown from read. reporting errors during destruction too is a
source of confusion for users since these errors are not usually
relevant and just the decompressor thread reporting short reads.
shorts reads during shutdown are not an error, they're expected.
fixes#1138
Change-Id: I7983ef0ed7722460f1d683bc8171a454b3d06e90
this includes tests that fail teardown when *skipped*, which confuses
the tap parser by reporting more tests than advertised in the header.
not sure what to do about this though, and it is better than nothing.
Change-Id: I4fc6badacbfa3b3d5f5f4a4ad76b177379bb4a30
pytest-tap didn't get it right in combination with xdist. we need to
replace the reporter *completely*, which is just incredibly weird :/
Change-Id: I2e35b9dc8ea4b24563b9b3306312778781eb674e
this requires some hacks to make error reports visible, but hey. it's
better than no progress reports at all, and successful runs no longer
spew huge amounts of useless log output onto the developer's terminal
Change-Id: I9e4766b2f825a4ec451e117eb4609db65f328785
not all operations that involve workers need to build. for example nix
copy --from calls instantiate thir worker with the *origin* store, not
the store they copy *into*. for binary cache stores this is not likely
to ever work if cgroups are enabled and the copy is run on normal user
credentials, even though no cgroups will ever be needed to copy things
fixes#1088
Change-Id: I065e29e1a1d0f58d81823609ef0701ff82cdd1d5
overridden-ness can be reset, and is usually reset such that setting
value taken from a config file are not considered as overriden. when
launching builtin builders we *do* want to send config file settings
changes to the builder, so we'll need one more getter method for it.
Change-Id: I861538a469121c77ebc1898a276439e6b756797d
diverted stores are only necessary when the logical store paths of
objects matter for the test itself, such as for derivation hashes,
substitution from golden sample nars, or actual tests of the store
diversion functionality. all other tests can use undiverted stores
to run, especially since only linux can build in diverted a store.
Change-Id: I62f0907bdef9961609af22b610195fcec54c1e57
StreamFdMessageReader reads message data *lazily*. if you don't access
all segments of the message before you close the underlying stream you
may find yourselv reading from something *very* different, and in this
case that something is The Void™. this causes reads to fail, writes on
the other side to fail to match, and finally our build launch to fail.
this does not lead to happy outcomes, so we will copy the full message
into a fresh new buffer *before* we try to access any of its contents.
fixes#1118
all-analysis-by: deprekated <kate@lix.systems>
Change-Id: I105540831fde855817194e9e539acf177f54a6f4
if a substituter is entirely offline and cannot be queries at all we
should not be failing if other substituters are configured. likewise
if a substituter goes offline after querying but before we try using
it we should attempt to fetch that path from some other substituter.
ideally we'd treat all substituters as a single entity instead of as
one store each, then have that single entity take care of fallbacks,
retries, error reporting, etc. that requires larger rewrites though.
fixes#1061
Change-Id: I9d8fc0544ff380bf017256e8fcc82823dc634f10
we don't use std::cerr often enough to want this, and since cerr flushes
after every operation anyway it never did anything useful to begin with.
Change-Id: Ia54be340826da5073e9a1786c463555f4f0e491c
kj exceptions get wrapped as Error instead of passed through as is.
luckily this also means that we can add context to them very easily
Change-Id: Icedab6c016f4434447dd38ba14138c102fc6149a
oops. we have to use init* here, not get*. get returns a discarding
builder for something that was not inited previously, but the linux
cause worked anyway because the first member is created by default!
Change-Id: I40f8a12a04eef2f4e3a80d1537ac9b975490a027
this allows us great flexibility in how children are launched (since the
actual launching is done by a separate executable), makes fork no longer
needed in the core codebase (outside of runProgram, anyway), and we even
get to use linux vfork to its full potential to decrease the launch cost
of sandboxes to a constant factor (previously it was O(#drvs + #deps) of
the build graph, which obviously goes to n² quickly if you are unlucky.)
Change-Id: I66e2d1b20242dc24d708666ef325fb8725bd9296
we can have a ProcessGroup for a pid that *should* be a process group
but hasn't gotten around to setting its pgid yet. in such cases we do
want to be killing the thing anyway, not shoot into the void and hope
the right thing falls over. so far this has not been a problem due to
a mixture of just not having done this and being slow enough to work.
Change-Id: I4e0e54513252d8e18256b9286b819bfa957d70dc
if we move this out of libstore and don't want to pull in all of libutil
with its myriad side effects std::filesystem is our best option, by far.
most notably we don't replace pathExists because std::filesystem::exists
behaves like `stat()` on symlinks, not like the `lstat()` as we require.
Change-Id: I1e488418dcabb33f2ebb73d8c3d1b43528aa51f1
eldritch horrors