Does anyone have issues with clang-with-thin-lto-ubuntu buildbot failing for practically every commit?

Hi,

I'm not really sure how the CI system works but I've consistently had failures
on that particular buildbot and it always seems on something completely unrelated
to any specific commit since the other bots work just fine when I look at the build
console and it seems to run fine (everything is green) and after committing it starts
failing for odd reasons.

In this particular case it was this:

/home/buildslave/ps4-buildslave1/clang-with-thin-lto-ubuntu/llvm.src/tools/clang/lib/
AST/ASTImporter.cpp:92:5: error: default label in switch which covers all enumeration
values [-Werror,-Wcovered-switch-default]
    default:
    ^
1 error generated.

Not entirely sure if I've been unlucky with timing or if there's an issue with it so
just wanted to check.

Thank you.
- Kristina

It sounds like that warning just needs to be fixed?

Bots are generally configured not to send email if they didn’t start /newly/ failing on the commit (or, unfortunately, the range of commits) in question. If you’re seeing that (an already failing bot send fail-mail to a newer commit that wasn’t involved in the change of state from green to red) - certainly something we should all look into.

Beyond that - some bots have longer cycle times which tends to mean larger commit ranges, which leads to a higher chance that any given cycle will fail (more changes, each with usually an independent/average chance of failure… etc). That’s unfortunate, but about the only solution is to find some way for the bot to run faster (so it can have smaller commit ranges between each run) or use more machines on that bot configuration (so each one kicks off a smaller range/runs in parallel with others) - if a bot has a really long cycle time such that the blame is rarely useful, I think it’s reasonable to push back a bit on the bot owner & perhaps they should be responsible for first level triage so as not to waste a lot of developers time looking at blames taht aren’t theirs.

As Zach said - looks like perhaps it picked up some other change along with yours that was, probably briefly, out of conformance with warnings?