Any buildbots building clang/llvm with gcc?

Hi,

Are there any buildbots that build clang/llvm with gcc and run tests nowadays?

When testing our out-of-tree target we are regularly building the compiler with both clang and gcc and lately we have noticed an increasing number of problems when using gcc which the rest of the community doesn't seem to notice?

E.g. for some days now at top-of-tree, if building with gcc 4.9.1 or 5.1.0 one test case fails with SEGV:

FAIL: Clang-Unit :: ASTMatchers/ASTMatchersTests/IsAnyPointer.MatchesObjcPointer (8218 of 25131)
******************** TEST 'Clang-Unit :: ASTMatchers/ASTMatchersTests/IsAnyPointer.MatchesObjcPointer' FAILED

I doubt any of the buildbots are running such recent versions of GCC - probably just whatever’s common/default on various distros like Ubuntu.

Adding a buildbot that covers them wouldn’t be a bad idea.

I doubt any of the buildbots are running such recent versions of GCC -
probably just whatever's common/default on various distros like Ubuntu.

This is true on my end.

Adding a buildbot that covers them wouldn't be a bad idea.

Good point. I'll add this to my list of TODOs on buildbots.

cheers,
--renato

Are there any buildbots that build clang/llvm with gcc and run tests
nowadays?

Many people use modern Linux distros which have GCC 5 on them. My Arch
has 5.3.0 and it works fine for me.

E.g. for some days now at top-of-tree, if building with gcc 4.9.1 or 5.1.0
one test case fails with SEGV:

Could it be a bad tree state? s this consistently broken over many
days of checkouts? Maybe it was just a bad patch that was reverted?

cheers,
--renato

I doubt any of the buildbots are running such recent versions of GCC -
probably just whatever's common/default on various distros like Ubuntu.

This is true on my end.

Adding a buildbot that covers them wouldn't be a bad idea.

Good point. I'll add this to my list of TODOs on buildbots.

Sounds good!

/Mikael

Hi,

Are there any buildbots that build clang/llvm with gcc and run tests
nowadays?

Many people use modern Linux distros which have GCC 5 on them. My Arch
has 5.3.0 and it works fine for me.

E.g. for some days now at top-of-tree, if building with gcc 4.9.1 or 5.1.0
one test case fails with SEGV:

Could it be a bad tree state? s this consistently broken over many
days of checkouts? Maybe it was just a bad patch that was reverted?

It's been consistently broken since the test case was added 4 days ago.

Now, it isn't a terribly long time but normally my experience is that "bad commits" gets reverted or fixed very fast, but things that we only discover when using GCC stays in tree longer.

/Mikael

The powerpc64 bots all use gcc. Most are using gcc 5.3.0 now but I think there's still one set that uses 5.1.

We build clang with gcc in Chromium a lot, but with an older gcc (4.8 at the moment).