Uh oh!
There was an error while loading. Please reload this page.
gh-95027: Ensure test runner uses utf-8:surrogateescape for communicating with subprocesses - #96669
gh-95027: Ensure test runner uses utf-8:surrogateescape for communicating with subprocesses#96669zooba wants to merge 1 commit into
Conversation
…municating with subprocesses
I tried to hand-apply this patch to installed 3.11.0rc1, but Lib/test/libregrtest/runtest_mp.py has changed too much. I tried copying the patched 3.12 libregrtest into 3.11, but that did not work either. I then ran the test with -j0 in updated main and got the same error as on the issue. Finally, I created a branch from this pr and ran the tests in paralled again and they ran to completion. So this patch appears to fix the issue I reported. (There were other problems -- see issue.) So from my viewpoint, it looks good. |
| env = dict(os.environ) | ||
| env['PYTHONIOENCODING'] = 'utf-8:surrogateescape' |
There was a problem hiding this comment.
I would prefer to test the Python "default" encoding, and so fix code which decodes test stdout instead.
There was a problem hiding this comment.
Python test suite worked well with any encoding for years without having to force an encoding, it was broken recently to support WASM.
There was a problem hiding this comment.
Then remove the encoding='utf-8' from the code that reads from the file.
But since this isn't testing, and is just IPC between test runners, I'd prefer to keep full fidelity in the data.
vstinner
commented
Oct 21, 2022
I wrote my PR #98492 Is this change still relevant? |
Uh oh!
There was an error while loading. Please reload this page.