llvm-reduce: Remove unsupported from bitcode uselistorder test

[arsenm]

This was disabled due to flakiness but I'm currently unable to reproduce.

I'm nervous the original issue still exists. However, I downgraded the tripped
assert in 8c18c25 to a warning since the same
assert can trigger for illegitimate reasons.

Fixes #64157

[arsenm]

• llvm-reduce: Remove unsupported from bitcode uselistorder test #134185 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[danilaml]

I remember hitting this assert not that long ago when reducing the test case for #131355 . Thankfully I was able to fallback to bugpoint, but in general it was easy to trigger eat when reducing miscompile caused by non-determinism in the compiler (which lack of uselists is only a small part of). If llvm-reduce can gracefully recover from this now, then great!

[arsenm]

I remember hitting this assert not that long ago when reducing the test case for #131355 . Thankfully I was able to fallback to bugpoint, but in general it was easy to trigger eat when reducing miscompile caused by non-determinism in the compiler (which lack of uselists is only a small part of). If llvm-reduce can gracefully recover from this now, then great!

It's not a graceful recovery, it's silently ignoring. If you can reduce produce a test that hits this, that would be great. Ignoring is the only thing we can do in the flaky test case, but if it's hit in a non-flaky case that's interesting

[danilaml]

@arsenm well, it was potentially flaky as the bug depended on the order of iteration of SmallPtrSet (actual bug was elsewhere but it was triggered by certain iteration orders) which depends on runtime values of pointers but in practice it was 100% reproducible o machines I've tried but failed in llvm-reduce. Maybe it had something to do with the way "counting chunks" was done (i.e. it wasn't the same as running fresh opt on IR). Not sure I have any tests that work with recent master and don't know if I'll be able to spot them if I hit this issue in the future. I guess I'll need to look out for the assert-turned-warning message (hopefully it wouldn't be drawn out by typical reduce log)?

[regehr]

LGTM!

[arsenm]

Unrelated windows failure, will just have to watch for other bots