This allows RemoteStore::addMultipleToStore() to free the Source
objects early (and in particular the associated sinkToSource()
buffers). This should fix#7359. For example, memory consumption of
nix copy --derivation --to ssh-ng://localhost?remote-store=/tmp/nix --derivation --no-check-sigs \
/nix/store/4p9xmfgnvclqpii8pxqcwcvl9bxqy2xf-nixos-system-...drv
went from 353 MB to 74 MB.
(cherry picked from commit cc838e8181)
First the motivation: I recently faced a bug that I assume is coming
from the topoSortPaths function where the GC was trying to delete a
path having some alive referrers. I resolved this by manually deleting
the faulty path referrers using nix-store --query --referrers. I sadly
did not manage to reproduce this bug.
This bug alone is not a big deal. However, this bug is
triggering a cascading failure: invalidatePathChecked is throwing a
PathInUse exception. This exception is not catched and fails the whole GC
run. From there, the machine (a builder machine) was unable to GC its
Nix store, which led to an almost full disk with no way to
automatically delete the dead Nix paths.
Instead, I think we should log the error for the specific store path
we're trying to delete, specifying we can't delete this path because
it still has referrers. Once we're done with logging that, the GC run
should continue to delete the dead store paths it can delete.
(cherry picked from commit ced8d311a5)
This fixes segfaults with nix copy when there was an error processing
addMultipleToStore.
Running with ASAN/TSAN pointed at an use-after-free with threads from
the pool accessing the graph declared in processGraph after the function
was exiting and destructing the variables.
It turns out that if there is an error before pool.process() is called,
for example while we are still enqueuing tasks, then pool.process()
isn't called and threads are still left to run.
By creating the pool last we ensure that it is stopped first before
running other destructors even if an exception happens early.
[ lix porting note: nix does not name threads so the patch has been
adapted to not pass thread name ]
Link: https://git.lix.systems/lix-project/lix/issues/618
Link: https://gerrit.lix.systems/c/lix/+/2355
(cherry picked from commit afac093b34)
In these trivial cases the final vector size (or lower bound on the size) is known,
so we can avoid some vector reallocations. This is not very important, but is just
good practice and general hygiene.
This is good practice to avoid pessimisations.
Left comments for the reasoning why ctors should be noexcept.
There are some tricky cases where we intentionally want throwing move ctors/assignments.
But those cases should really be reviewed, since some of those can be replaced
with more idiomatic copy/move-and-swap.
`auto &&` and `T &&` are forwarding references and can be
either lvalue or rvalue references. Moving from universal references
is incorrect and should not be done.
Moving from integral or floating-point values is pointless and just
worsens debug performance.
the default int64_t max was still overflowing for me, when this was dumped as json (noticed during building the manual).
So making 0, the default and define it as "no warnings" fixes the situtation.
Also it's much more human-readable in documentation.
This works because the `builder` and `args` variables are only used
in the non-builtin code path.
Co-Authored-By: Théophane Hufschmitt <theophane.hufschmitt@tweag.io>
Otherwise, if checkInterrupt() in any of the supported store operations
would catch onto a user interrupt, the exception would bubble to the thread
start and be handled by std::terminate(): a crash.
... and remove a few unused arguments.
This adds pkg-config to a two or three packages that don't use it,
but we shouldn't let that bother us. It's like our personal stdenv.
If you have the Nix store mounted from a nonlocal filesystem whose
exporter is not running as root, making the directory mode 000 makes it
inaccessible to that remote unprivileged user and therefore breaks the
build. (Specifically, I am running into this with a virtiofs mount using
Apple Virtualization.framework as a non-root user, but I expect the
same thing would happen with virtiofs in qemu on Linux as a non-root
user or with various userspace network file servers.)
Make the directory mode 500 (dr-x------) to make the sandbox work in
this use case, which explicitly conveys our intention to read and search
the directory. The code only works because root can already bypass
directory checks, so this does not actually grant more permissions to
the directory owner / does not make the sandbox less secure.
This caused nlohmann/json.hpp to leak into a lot of compilation units,
which is slow (when not using precompiled headers).
Cuts build time from 46m24s to 42m5s (real time with -j24: 2m42s to
2m24s).
These versions are more than 3 years old and were very early in the
existence of CA derivations support (which was and is experimental),
so they're unlikely to still exist in the real world. So let's get rid
of support for them.
