Uh oh!
There was an error while loading. Please reload this page.
[3.12] gh-110276: No longer ignore PROFILE_TASK failure silently (#110295) - #111950
[3.12] gh-110276: No longer ignore PROFILE_TASK failure silently (#110295)#111950vstinner wants to merge 1 commit into
Conversation
…on#110295) (cherry picked from commit 6ab6040)
vstinner
commented
Nov 10, 2023
We should not ignore test failure when running the PGO build: see the new failure #111929 impacting 3.11 and 3.12 branches. |
vstinner
commented
Nov 10, 2023
Once merged, this change should be backported to the 3.11 branch once #111949 is merged. |
vstinner
commented
Nov 10, 2023
cc @AlexWaygood |
vstinner
commented
Nov 10, 2023
vstinner
commented
Nov 11, 2023
Oh, the initial issue was #101634 |
vstinner
commented
Nov 12, 2023
@gpshead: 3.11 and 3.12 are also affected. Are you fine with my change? |
gpshead
commented
Nov 13, 2023
yeah backporting that makes sense. |
Yhg1s
commented
Nov 13, 2023
No, I would rather not backport this change. The job of the profiling run isn't to check whether tests pass or not. It's to exercise enough of Python and the standard library to make a reasonable profile. A test failure doesn't mean that didn't happen. Even not running any tests doesn't mean there is no reasonable profile. I don't think it's enough of a problem to warrant breaking existing 3.12 build processes. Making the profile task test failures fatal in 3.13 is fine, but I'd rather not disrupt anyone's existing 3.12 builds and make upgrading to 3.12.1 that much harder. |
vstinner
commented
Nov 13, 2023
@AlexWaygood: Python 3.12 has spoken, I close this PR ;-) |
vstinner
commented
Nov 13, 2023
If tomorrow, this issue becomes a major issue, maybe we can consider adding an option to regrtest to return a non-zero exit code if |
(cherry picked from commit 6ab6040)
test_unicodealways fails during a PGO-optimised build (because it does not exist) #110276📚 Documentation preview 📚: https://cpython-previews--111950.org.readthedocs.build/