This is an archive of the discontinued LLVM Phabricator instance.

[sanitizers] fix print-module-map test on linux
ClosedPublic

Authored by aralisza on Mar 5 2021, 4:27 PM.

Details

Summary

Looks like the default options for halt_on_error are different between linux and mac. set it to 0 in the test so the behavior is the same on both platforms.

rdar://75110847

Diff Detail

Event Timeline

aralisza requested review of this revision.Mar 5 2021, 4:27 PM
aralisza created this revision.
Herald added a project: Restricted Project. · View Herald TranscriptMar 5 2021, 4:27 PM
Herald added a subscriber: Restricted Project. · View Herald Transcript
aralisza edited the summary of this revision. (Show Details)Mar 5 2021, 4:38 PM
delcypher accepted this revision.Mar 5 2021, 4:38 PM

LGTM.

This revision is now accepted and ready to land.Mar 5 2021, 4:38 PM

looks like the Debian pre-merge checks passed. not going to wait for windows since this test isn't run on windows

This revision was landed with ongoing or failed builds.Mar 5 2021, 4:41 PM
This revision was automatically updated to reflect the committed changes.

looks like the Debian pre-merge checks passed. not going to wait for windows since this test isn't run on windows

I guess this is not a solution.
It worked on Debian before, it's something about PPC

"ABORTING" in case of ubsan is result of UndefinedBehaviorSanitizer:DEADLYSIGNAL, not ubsan report itself.
Somehow it didn't trigger signal on PPC, so hald on error will not help

However "abort_on_error=1" may help as it makes print ABORT even before signal.
abort_on_error=1 is default on Darwin/Android and 0 on the rest

What do you think if I just disable the test on power pc? I'm not sure I'm the best person to be debugging this.

Never mind, I see that you got this, thanks!