[test][IRGen] Fix check on local_extern test. #33816
Conversation
|
@swift-ci please smoke test and merge |
|
CC: @xedin |
|
@swift-ci please test Windows platform |
|
@3405691582 seems that its not enough? |
|
@3405691582 I think we need to revert; this is breaking Foundation builds on Windows too as it no longer correctly is processing typedefs. |
|
Nope, false alarm there. |
|
I suspect this is another rigid match problem, but the CI is not showing me the full IR from that unit test so I can't confirm. Are you able to grab the emitted IR and show it to me, or let me know if there's a way to see it simply? |
|
Not currently, there is something else that is preventing me from building, but the CI hosts seem okay (I think that the SDK version might be exposing another issue locally). |
|
As an experiment, I'm going to try removing the expectation on the function case, just to see if the other cases are as expected. Can you rerun the CI on Windows? |
3405691582
force-pushed
the
FixLocalExtern
branch
from
41a154a to
4c6aa9c
Compare
|
@swift-ci please test Windows platform |
3405691582
force-pushed
the
FixLocalExtern
branch
from
4c6aa9c to
e1b041e
Compare
|
ugh, I cannot into git. You might have to do that again. |
|
@swift-ci please test Windows platform |
Windows IR seems to include a `dso_local` annotation, which the check didn't expect and caused a failure in swiftlang#33813. Added a regex check to ensure the test passes with or without the annotation.
3405691582
force-pushed
the
FixLocalExtern
branch
from
e1b041e to
1acecd1
Compare
|
That appears to have worked. Therefore, I think a valid approach for now is to create an amended version of the test for Windows and look into examining IR or figuring out a different test mechanism at a later date. Do you agree, or should I try something else? |
|
Sounds good to me, lets fix forward if we can. |
|
@swift-ci please smoke test |
|
@swift-ci please smoke test Linux platform |
In swiftlang#33816, we forked the local_extern test omitting the check for the extern function, because it wasn't obvious what the match was. It might actually just be missing the dso_local annotation again for the function. Let's try that.
Windows IR seems to include a
dso_localannotation, which the checkdidn't expect and caused a failure in #33813. Added a regex check to
ensure the test passes with or without the annotation.