The patch has been committed on your behalf.

Do you need a help with the commit?

The staging is ready for you. Please feel free to commit.

LGTM

Please feel free to connect to the staging.

Frankly, I don't like the idea of allowing failing tests, definitely not in production. I see what you want to achieve with this, though.

LGTM

Did you analyze what's the most often reason of git fetch failures? Shall we also set clobberOnFailure to escalate a clone?

LGTM

If this is fine with you, let's finish this patch first, stage the builder to make sure it is reliably green, then you could move it to the production build bot and add to the release_builders.py.

Also I had questions on about if I needed to rebuild llvm, clang and lld-devel to test; is it possible to use what was built beforehand by previous tests on the same commit ?

And I was even looking at it there!

But please wait for the patch Andrzej mentioned before committing.

LitTestCommand parses LIT logs only and is not intended to parse anything else, AnnotatedCommand parses annotations only and is not intended to parse anything else. NinjaCommand could build (build logs), test (LIT logs), install (install logs).

Hello Brice,

Before committing, could you make sure you are based on the current UnifiedTreeBuilder and LLVMBuildFactory, please?
If I remember the context correctly, you may need to specify enable_runtimes="auto", since the default behavior has been changed in https://reviews.llvm.org/rZORGf7b888457641 per request.

I would not characterize the ninja invocation doing the build as a "single command".

So, your builder does not send anything to TTY for more than 20 minutes during a normal regular build. Meaning that a single commands takes longer than 20 minutes to execute.

Almost forgot to mention.

LGTM

Sorry for the delay.

Anyway, the default has been changed in https://reviews.llvm.org/rZORGf7b888457641 to not use LLVM_ENABLE_RUNTIMES unless explicitly requested.
That should address your concern.
Thanks again, Michael.

This is why I asked if you know what builders should build without using LLVM_ENABLE_RUNTIMES but now they don't.
The change we are discussing does not necessarily mean that LLVM_ENABLE_RUNTIMES is always engaged.

To make sure I'll check the fix, could you point me to a particular builder or builders where you see the problem, please?

Hello Michael,

Thanks for the patch, Michael!

Please see my comments inline.

This patch introduced a bug.

Looks good with a nitpick.

LGTM

Looks good with a nitpick. Please see my comment inline.

Thanks! Looks good now.

Yep. The raw diff looks good.

Sorry for reverting your patches, but I do not see a quick and easy way to fix the problems it has introduced.

This patch has broken zorg. I reverted the both this one and D99097 as they seems closely related, and D100431 fixes D99097.

Looks good with a nit pick.
Could you clean the patch off of the unrelated changes, please?

Could you also clean up config/workers.py and config/status.py from the unused now workers, please?

No. Cuda builders got that in the code base as a temporary solution while they are working on their annotated script. Exactly the same arguments as yours. And I see that being confusing, and don't think it worth it after all.

Thanks for updating the LLVM Developer Policy, Christian.
It looks good with a couple nit picks.

LGTM

LGTM

Everything else looks Ok.

Looks fine.

LGTM

Hm. This is weird.
I see a wrong patch assigned to this review now.,

Thanks, Albion!
Looks much better now.

LGTM

LGTM

LGTM

Almost there.
One problem to fix.

I have added the support you need to UnifiedTreeBuilder.

LGTM

There are few comments for the buildbot part.

LGTM

LGTM

LGTM

Looks good with a couple of nit picks.
Please see my comments in line.

Looks good with a couple nit picks.
Please see my comments in line.

It is good with a couple of small things to address.

It seems this patch needs more attention.
Please see my comments above.

I'm not sure I follow what are you trying to achieve by changing the src_root.

Christian, do you need a help with committing this patch? You should be good if you have LLVM commit permissions, but please let me know if you want me to commit.

LGTM

LGTM.

LGTM

LGTM