reland: Emit coverage mappings for all modules #32352
Conversation
|
Do you have an SR for the XFAIL on Windows? |
|
@swift-ci smoke test |
|
added to the description, also happy to add in a comment there if you'd prefer! |
|
That's all I need for now! To aid your debugging, have you tried marking both files as |
|
Interesting, I'm still working on getting a windows VM setup to actually debug at all 😞 but that's a good thing I can try! |
|
seems like my XFAIL didn't work, does it have to be at the top or something? I copy pasted from some other test and that's the only major difference I see |
|
Try using |
keith
force-pushed
the
ks/reland-emit-coverage-mappings-for-all-modules
branch
from
eabf16b to
2299968
Compare
|
I tested your primary-file idea on macOS and it still works as expected, I pushed that here and removed the XFAIL, do you mind re-running the tests so we can see if that works for windows too (I'm still building on my windows VM) |
|
looks like we commented around the same time, I can push that if this option doesn't work while I continue to debug it |
|
@swift-ci smoke test |
|
hrm this windows failure looks like a flake |
|
@swift-ci smoke test Windows |
|
Yay! It looks like your idea worked! Thanks a ton! |
|
Ah crap, it looks like while the test did pass in this configuration, it's because it's not actually validating the right behavior. If I revert my actual change and run with the |
keith
force-pushed
the
ks/reland-emit-coverage-mappings-for-all-modules
branch
from
2299968 to
5c8ec2e
Compare
|
I've reverted that change so that it fails again, and will continue to debug a bit, I now have a windows environment setup and I can't reproduce the CI failure 😞 |
|
@swift-ci please test Windows platform |
|
was hoping the behavior might differ if I was trying to rebuild on the same commit |
|
@swift-ci please test Windows platform |
This reverts commit 499ed05. This changes the num-threads passed to 1 to avoid a multithreaded output issue on windows
keith
force-pushed
the
ks/reland-emit-coverage-mappings-for-all-modules
branch
from
c941648 to
76ca05f
Compare
|
@swift-ci please test Windows platform |
|
Yippee it passed! |
|
@swift-ci please test |
|
Build failed |
|
Build failed |
|
Thanks again everyone for your help! |
This reverts commit 499ed05.
The change here is to pass
-num-threads 1to avoid an issue on windows where the output of both modules was interleaved. This still goes down the multithreaded codepath so it still validates the fix.I filed https://bugs.swift.org/browse/SR-13016 for future investigation around this multithreaded issue
#32216 (comment)