bugzilla complaining about "unexpected error"

Hi!

Recently I've seen errors like this in Bugzilla when I've added comments etc to trouble reports:

An unexpected error occurred. This could be a temporary problem, or some code is behaving incorrectly. If this problem persists, please email this page to llvm-admin@lists.llvm.org with details of what you were doing at the time this message appeared.

URL: https://bugs.llvm.org/process_bug.cgi?rep_platform=PC&product=libraries&bug_status=NEW&classification=Unclassified&delta_ts=2021-09-27%2007%3A17%3A58&deadline=&defined_bug_ignored=1&keywords=&token=1632826824-LaTofBNN2ThrLhV2WPxFfYGudi4QLRucdsbG96dubu0&blocked=&bug_severity=enhancement&resolution=&component=Loop%20Optimizer&priority=P&cf_fixed_by_commits=&estimated_time=0.0&dup_id=&assigned_to=unassignedbugs%40nondot.org&bug_file_loc=&short_desc=opt%20-passes%3D'loop-mssa(canon-freeze%2Cloop-rotate)%2Cfunction(bounds-checking)'%20fails%20with%20Assertion%20`isAvailableAtLoopEntry(OrigStartMinusStride%2C%20L)%20%26%26%20"Must%20be!"'&version=trunk&op_sys=Linux&dependson=&newalias=&addselfcc=on&comment=Just%20some%20additional%20info. It%20also%20fails%20when%20using%20different%20loop%20pass%20managers%20like%20this%3A %20%20-passes%3D'loop-mssa(canon-freeze)%2Cloop-mssa(loop-rotate)%2Cfunction(bounds-checking)' If%20we%20invalidate%20scalar-evolution%20after%20canon-freeze%20like%20this %20%20-passes%3D'loop-mssa(canon-freeze)%2Cinvalidate<scalar-evolution>%2Cloop-mssa(loop-rotate)%2Cfunction(bounds-checking)' then%20it%20works%20fine%20(and%20also%20if%20we%20split%20this%20into%20running%20opt%20twice%20with%20canon-freeze%20in%20the%20first%20run%20and%20the%20rest%20of%20the%20pipeline%20in%20a%20subsequent%20run).%20So%20maybe%20it%20is%20canon-freeze%20that%20isn't%20updating%20scalar-evolution%20analysis%20correctly.&id=51981&tag=&work_time=0&see_also=&remaining_time=0.0
undef error - This shouldn't happen at /usr/share/perl/5.18/Text/Wrap.pm line 84.

Those are a bit annoying, but the tickets at least seem to get updated anyway.
Although, there are some suspicion that some email-notifications might go missing (at least one time it has happened that the reporter of a bug did not receive an email when I added comments to a PR).

This time it was related to PR51981, and I simply got the failure message above when adding "Comment 1".

According to Mikael (who wrote the ticket):
- he got a similar message when writing the ticket
- he did not receive any email notification when I wrote my comment
- he got a similar message when adding himself as CC, and when doing that he also got an error message like above

Regards,
Björn

I got the same error message right now (but for anopther PR than last time):

An unexpected error occurred. This could be a temporary problem, or some code is behaving incorrectly. If this problem persists, please email this page to llvm-admin@lists.llvm.org with details of what you were doing at the time this message appeared.

URL: https://bugs.llvm.org/process_bug.cgi?classification=Unclassified&assigned_to=unassignedbugs%40nondot.org&comment=The%20same%20assert%20can%20also%20be%20seen%20with%3A %20opt%20-passes%3D"cgscc(devirt<4>(inline))%2Cfunction(slp-vectorizer)%2Ccgscc(devirt<4>(function(sroa)))"%20-o%20%2Fdev%2Fnull%20bbi-53898_2.ll%20-aa-pipeline%3Ddefault In%20this%20case%20it%20starts%20happening%20with%20commit%2078d4096d037%3A %20%20%20%20[LazyCallGraph]%20Add%20invalidate%20method. %20%20%20%20 %20%20%20%20Summary%3A%20Add%20invalidate%20method%20in%20LazyCallGraph. %20%20%20%20 %20%20%20%20Reviewers%3A%20chandlerc%2C%20silvas %20%20%20%20 %20%20%20%20Subscribers%3A%20hiraditya%2C%20sanjoy.google%2C%20llvm-commits %20%20%20%20 %20%20%20%20Tags%3A%20%23llvm %20%20%20%20 %20%20%20%20Differential%20Revision%3A%20https%3A%2F%2Freviews.llvm.org%2FD72817 In%20a%20later%20commit%2C%206699029b67b%2C%20-aa-pipeline%3Ddefault%20was%20made%20the%20default%2C%20so%20after%20that%20it%20can%20be%20reproduced%20with%20just %20%20opt%20-passes%3D"cgscc(devirt<4>(inline))%2Cfunction(slp-vectorizer)%2Ccgscc(devirt<4>(function(sroa)))"%20-o%20%2Fdev%2Fnull%20bbi-53898_2.ll&newalias=&id=51946&bug_status=NEW&remaining_time=0.0&component=Interprocedural%20Optimizations&estimated_time=0.0&rep_platform=PC&bug_severity=enhancement&dependson=&op_sys=Linux&version=trunk&delta_ts=2021-09-23%2005%3A35%3A25&priority=P&work_time=0&blocked=&short_desc=opt%20-passes%3D'cgscc(require<fam-proxy>)%2Cipsccp%2Cscc-oz-module-inliner'%20fails%20with%20Assertion%20`(E%20||%20!FunctionPass)&product=libraries&deadline=&bug_file_loc=&token=1633421733-dXmIzxuysImXJAUDZgfe7YM2FqAEtboFOSu-4lNkkdQ&tag=&see_also=&cf_fixed_by_commits=&defined_bug_ignored=1&dup_id=&keywords=&resolution=
undef error - This shouldn't happen at /usr/share/perl/5.18/Text/Wrap.pm line 84.

Traceback:

I got it when adding a comment, and the comment did get added even with the above error, but I don't know if mail notifications were sent out of people on the issue CC list or not.

/Mikael

And to add, when I tried attaching an ll-file to the same PR I got:

Internal Server Error

The server encountered an internal error or misconfiguration and was unable to complete your request.

Please contact the server administrator at llvm-admin@lists.llvm.org to inform them of the time this error occurred, and the actions you performed just before this error.

More information about this error may be available in the server error log.

Also in this case, the file did indeed get attached, but no idea if any on the CC list got notified.

/Mikael

I haven't seen any email notifications related to your updates to PR51946 today. That is quite annoying.

/Björn

Looks like there’s an upstream bug report regarding “undef error - This shouldn’t happen at /usr/share/perl/5.18/Text/Wrap.pm line 84.”:

https://github.com/Perl/perl5/issues/13447

https://rt.cpan.org/Public/Bug/Display.html?id=124024

If I had to guess, it’s triggered for Mikael’s issues due to the bug title style, which have very very long “words” (e.g. “-passes=‘loop-mssa(canon-freeze,loop-rotate),function(bounds-checking)’”).

As a workaround, I’d suggest not putting the reproduction command-lines in the bug title.

Aha, yes maybe that's it. I'll try to keep the length down then and see.

Thanks for the suggestion,
Mikael