Fixed Bug 41532 - update_test_checks does not actually handle block names. Also rerun reassociate-after-unroll.ll.
I will address the issue in the script at a later date and try to provide a fix for PR41532, but for now I have this test case failing unnecessarily in an out-of-tree target. So I’d like to proceed with the interim fix to the test case for the time being.
I cannot think of any clean way to identify labels, there are three ways that label can be:
- at the start of a basic block.
- follow by "label".
- incoming blocks in phinode.
@MaskRay Do you feel strong to limit the change only to label? If so, can you give me some guidance on how to do it?
I'm not sure I see the value in this change in either implementation. Changing block names seem to be a pretty niche issue (haven't ever run into this myself) and I don't think handling them in update_test_checks is worth the churn it will cause.
@nikic Thanks for the suggestion. I personally ran into the issue that reassociate-after-unroll.ll failed due to the different in basic block name, after modifying the O2 pass pipeline. And modifying the script, was the suggested way to solve the failure. I am willing to put this change behind a flag if that's preferred, but it maybe hard for user to know when to use the flag, or even know the existence of it.