New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
test_context and test_asyncio crash on Windows 7 #76822
Comments
http://buildbot.python.org/all/#/builders/58/builds/498 0:36:25 [252/414/1] test_context crashed (Exit code 3) -- running: test_largefile (239 sec) Current thread 0x00000944 (most recent call first): Current thread 0x00000944 (most recent call first): 0:46:50 [306/414/2] test_asyncio crashed (Exit code 3) -- running: test_io (487 sec) Current thread 0x00000d24 (most recent call first): Current thread 0x00000d24 (most recent call first): Note: the commit 0bad4d6 doesn't fix these crashes. (The running build 399 hash this fix, and test_context does still crash.) |
test_asyncio started to crash at build 491: http://buildbot.python.org/all/#/builders/58/builds/491 This build contains the commit f23746a: "bpo-32436: Implement PEP-567". I'm not 100% sure that it's the cause of the regression. Since test_context added by this commit also crash, it really smells this the cause of the regression. |
It's not a regression, the newly added code doesn't work on some Windows buildbot... looking at it. |
Some hints. Compiler warnings: ..\Python\hamt.c(625): warning C4018: '<': signed/unsigned mismatch [D:\cygwin\home\db3l\buildarea\3.x.bolen-windows7\build\PCbuild\pythoncore.vcxproj] |
No, these are trivial uint32_t/Py_SIZE: |
I suspect this is a compiler bug. I can't reproduce it on my windows 7 virtual machine and on AppVeyor. It's not related to HAMT, btw, as there is a multitude of HAMT-specific tests that all pass. The crash is specifically in context.c, and the code there is pretty trivial. |
It may be interesting to add the compiler version to test.pythoninfo. It would be simpler to compare Windows buildbots to identify a bug. Currently, the only info is: "sys.version: 3.7.0a4+ (heads/master:9d411c1, Jan 23 2018, 15:13:42) [MSC v.1900 32 bit (Intel)]" Some environment variables are also dumped: os.environ[VS100COMNTOOLS]: C:\Program Files\Microsoft Visual Studio 10.0\Common7\Tools\ In the compile step, I see lines like: "Microsoft (R) Build Engine version 14.0.24730.2" I recall also issues with outdated ucrtbase DLLs. Maybe we can also dump the version of such DLL? |
In an off-list conversation with David, it was confirmed that the buildbot in question runs on an old hardware without SSE4.2 support. I decided to simply stop using native popcount instructions as there's no detectable performance difference when using them vs using the fallback code. See #5292 for details. |
I've pushed a commit that should fix the buildbot. Please reopen if it doesn't. |
test_context and test_asyncio do'nt crash anymore on this buildbot, thanks Yury. test_asyncio still fail on this buildbot, I created a new issue: bpo-32645, test_asyncio: TLS tests fail on "x86 Windows7" buildbot. |
Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.
Show more details
GitHub fields:
bugs.python.org fields:
The text was updated successfully, but these errors were encountered: