Spurious access violation on Windows

[alexcrichton]

http://buildbot.rust-lang.org/builders/auto-win-gnu-64-opt/builds/4086/steps/test/logs/stdio
https://gist.github.com/alexcrichton/847ad18aec0a73d342838e5d8ed683a7
#33086 (comment)

---- [run-pass] run-pass\macro-crate-does-hygiene-work.rs stdout ----

error: auxiliary build of "C:/bot/slave/auto-win-gnu-64-opt/build/src/test/auxiliary/macro_crate_test.rs" failed to compile: 
status: exit code: 3221225477
command: PATH="C:\bot\slave\auto-win-gnu-64-opt\build\obj\x86_64-pc-windows-gnu/stage2/bin;C:\bot\slave\auto-win-gnu-64-opt\build\obj\x86_64-pc-windows-gnu\stage2\bin;C:\bot\slave\auto-win-gnu-64-opt\build\obj\x86_64-pc-windows-gnu\llvm\Release\lib;C:\mingw-w64\x86_64-4.9.1-win32-seh-rt_v3-rev1\mingw64\bin;C:\Python27;C:\msys64\mingw64\bin;C:\msys64\usr\bin;C:\Windows\system32;C:\Windows;C:\Windows\System32\Wbem;C:\Windows\System32\WindowsPowerShell\v1.0;C:\msys64\usr\bin;C:\python27;C:\python27\scripts;C:\program files (x86)\inno setup 5;C:\program files (x86)\CMake\bin;C:\Program Files (x86)\Windows Kits\8.1\Windows Performance Toolkit;C:\Program Files\Microsoft SQL Server\110\Tools\Binn;C:\Program Files (x86)\Microsoft SDKs\TypeScript\1.0" x86_64-pc-windows-gnu/stage2/bin/rustc.exe C:/bot/slave/auto-win-gnu-64-opt/build/src/test/auxiliary/macro_crate_test.rs -L x86_64-pc-windows-gnu/test/run-pass-fulldeps/ --target=x86_64-pc-windows-gnu --crate-type=dylib -L x86_64-pc-windows-gnu/test/run-pass-fulldeps\macro-crate-does-hygiene-work.stage2-x86_64-pc-windows-gnu.run-pass.libaux -C prefer-dynamic --out-dir x86_64-pc-windows-gnu/test/run-pass-fulldeps\macro-crate-does-hygiene-work.stage2-x86_64-pc-windows-gnu.run-pass.libaux --cfg rtopt -C rpath -O -L x86_64-pc-windows-gnu/rt
stdout:
------------------------------------------

------------------------------------------
stderr:
------------------------------------------

------------------------------------------

thread '[run-pass] run-pass\macro-crate-does-hygiene-work.rs' panicked at 'explicit panic', C:/bot/slave/auto-win-gnu-64-opt/build/src/tools/compiletest/src\runtest.rs:1595
note: Run with `RUST_BACKTRACE=1` for a backtrace.


failures:
    [run-pass] run-pass\macro-crate-does-hygiene-work.rs

test result: FAILED. 56 passed; 1 failed; 2 ignored; 0 measured

The exit code 3221225477 is 0xC0000005 in hex which correspond to an access violation. Basically this is a spurious segfault on Windows (I think).

I've seen this from time to time, but unfortunately it doesn't seem to reproduce on the same test so far nor have I ever seen it locally. Seems bad though!

[alexcrichton]

Failures like #33072 (comment) (full) make me think that this is some crazy thing because rm segfaulted. I could be wrong though.

[alexcrichton]

One possible explanation may be OOM, although I have nothing to support that beyond a vague mention in a thread that is itself only vaguely related to this. Looking at a build which failed for this there were 3 builds on one Windows machine (the max concurrent) where one finished, another died, and the last was killed all at about the same time. The one which was killed was compiling LLVM (one file at a time), the one which died was in pretty tests. Neither of those are really that memory intensive and the machines have 16GB of memory which should in theory be more than enough...

Something to keep an eye out for I guess?

[matthiaskrgr]

If this is the same as mentioned here https://randomascii.wordpress.com/2018/02/25/compiler-bug-linker-bug-windows-kernel-bug/ , I wonder if we could implement a workaround somehow.

[steveklabnik]

It's been a while; are we still seeing this?

[Brooooooklyn]

I saw the same issue while implementing a NodeJS napi library: napi-rs/napi-rs#297

It can reproduce in https://github.com/napi-rs/package-template

0. choco install node-lts -x86 -y -f
1. refreshenv
2. checkout napi-1-pre branch
3. yarn install
4. yarn build
5. node simple-test.js

The reproduce step could be found here: https://github.com/napi-rs/package-template/blob/napi-1-pre/.github/workflows/CI.yaml#L189

You can also see Github actions output from napi-rs/package-template#86

[ChrisDenton]

I'm no longer able to reproduce the original issue nor the other case (please correct me if you can). It's not clear if they're even related as a spurious access violation can be caused by any number of different things.

I'm minded to close this in favour of more specific issues. If anyone is experiencing 0xC0000005 (aka STATUS_ACCESS_VIOLATION) it would be particularly useful to open a new issue with a minimal (if possible) reproducible example using nightly.

[jieyouxu]

Triage: I'm inclined to close this as well due to lack of repro and overly generic failures, in favor of more specific issues if people are still running into them.