Slew of new failures in Linux, caused by llvm

Subprojects LLDB
#1

Hi Rafael,

http://lab.llvm.org:8011/builders/lldb-x86_64-debian-clang/builds/3326

Shows about 20 new test failures. I’ve narrowed it down to r228886

http://llvm.org/viewvc/llvm-project?revision=228886&view=revision

Can you take a look and/or give me some idea what to look for?

Thanks,

Vince

#2

We should revert this if it’s definitely the culprit. Otherwise it’s impossible for to detect another patch later which also causes a regression. It’s easy to fix and resubmit later

#3

It doesn’t revert cleanly

#4

Hmm I checked the logs, that file hasn’t been touched since that revision. What’s unclean about the revert? Compile failure, or other test failure?

#5

IIRC I used

llvm$ svn diff --revision 228885:228886 > ~/break.diff

llvm$ patch -p0 -R < ~/break.diff

Some hunks failed. Could’ve messed something up.

#6

Yea theres 886, 888, 889, and 890. I think all would have to be reverted. Is anyone on #llvm IRC that can help?

#7

No one has even replied to the commit thread? that’s where this kind of discussion should took place.

#8

I just joined llvm-dev today. Next time I’ll have the commit thread. =)

#9

llvm-commits you mean?

You can reply via the archives or one of the other services?

In any event, you prabably want to subscribe to but auto-filter all of
cfe-coommits and llvm-commits sense lldb can be broken by commits to either.

#10

Sorry I missed that. It is strange that only lldb tests found the issue, but by all means, please revert until green.

I don’t have access to a computer at the moment. I will be back on Tuesday.

Sorry once more.

Yea theres 886, 888, 889, and 890. I think all would have to be reverted. Is anyone on #llvm IRC that can help?

#11

I'll revert the commit series for you in a bit Rafael. No stress.

#12

Thanks Chandler!

#13

Backed out in r229080. Will watch to make sure this actually helps the bots and re-instate otherwise.