LGTM

Hello Endre,

Thanks for fixing this, Vlad!

LGTM

Thanks for adding new builders, Vlad!

LGTM

LGTM

Thanks for cleaning this!

This version would keep the llvm-clang-x86_64-expensive-checks-win bot happy.
Thanks, Tim!

This makes things better, but did not resolve the problem completely.
The following 2 tests are still failing:

********************
Failing Tests (2):
    LLVM :: CodeGen/MIR/AMDGPU/machine-function-info-no-ir.mir
    LLVM :: CodeGen/MIR/AMDGPU/machine-function-info.ll

Both failures has the same reason behind:

>c:\ps4-buildslave2\llvm-clang-x86_64-expensive-checks-win\build\bin\llc.exe -run-pass=none -verify-machineinstrs C:\ps4-buildslave2\llvm-clang-x86_64-expensive-checks-win\build\test\CodeGen\MIR\AMDGPU\Output\machine-function-info.ll.tmp.mir -o -
error: C:\ps4-buildslave2\llvm-clang-x86_64-expensive-checks-win\build\test\CodeGen\MIR\AMDGPU\Output\machine-function-info.ll.tmp.mir:114:21: expected a named register
  scratchRSrcReg:  ''
                    ^

Hello Leandro,

Thanks, Michał, for making this change.

I made the LLDBBuilder.getLLDBScriptCommandsFactory change for you.

I'd prefer seeing getLLDBScriptCommandsFactory changed to use LLVMBuildFactory with the properly defined dependencies, rather than another unique scheduler for one or very few builders.
If you would use LLVMBuildFactory , build master will take care of the correct scheduling the changes for you.
You may take a look at zorg/buildbot/builders/FuchsiaBuilder.py as an example of how this could be done easy.

Thanks for making this more clean, Jan!
Since your bot is on the silent master yet, I can push the change there as soon as you commit.

LGTM

LGTM

Thanks, Petr!
LGTM

Thanks for working on this, Alexandre!

Could you clean off the unused imports, please?
You could test how the factory works on the silent master. I'll help if any debugging would be needed.

It is better to use the zorg.buildbot.process.factory.LLVMBuildFactory instead of buildbot.process.factory.BuildFactory(). Zorg would assign it to the correct scheduler automatically in this case.

Hello Petr,

Hi Stella,

Please revert r345309 and propose a smaller change.

The latest changes look too invasive for the purpose. And it seems confusing to have two parameters, both supposedly controlling the Visual Studio selection, especially when one could be silently ignored in some cases.

I see that in the code. The question was if you actually tested it with something older than VS 2015.
Anyway, LGTM.

Stella,
Did you try the change with VS other than 2017?

LGTM

LGTM

LGTM

Hello Krzysztof,

LGTM

LGTM

LGTM

LGTM

LGTM

LGTM

LGTM.
This seems a proper GTest way, indeed.
Just for the record, the warning was reported by clang 5.0.1.

LGTM

LGTM

LGTM.
There is a cosmetic issue, but your change seems in line with the rest of the file, so, not sure if you would want to do this just for 2 ninja commands you are adding, or address it everywhere in this file as a separate patch.

LGTM