$ clang --version Ubuntu clang version 3.6.0-2ubuntu1 (tags/RELEASE_360/final) (based on LLVM 3.6.0) Target: x86_64-pc-linux-gnu Thread model: posix build error: ------------- pixman-mmx.c:100:20: error: constraint 'K' expects an integer constant expression : "y" (__A), "K" (__N) ^~~ related bug reports: - https://github.com/Homebrew/homebrew/issues/41056 - http://lists.freedesktop.org/archives/pixman/2015-October/004080.html Let's disable MMX optimizations to make clang happy.
Created attachment 265859 [details] Patch
If this is a problem only for clang, would it be possible to detect it somehow from jhbuildrc and add the autogen command line argument there?
(In reply to comment #2) > If this is a problem only for clang, would it be possible to detect it > somehow from jhbuildrc and add the autogen command line argument there? I don't know if it is possible, but maybe it is easier to patch configure.ac to disable mmx. It has some check for this, but it seems it doesn't work properly. I'll check it later.
(In reply to comment #3) > (In reply to comment #2) > > If this is a problem only for clang, would it be possible to detect it > > somehow from jhbuildrc and add the autogen command line argument there? > > I don't know if it is possible, but maybe it is easier to patch > configure.ac to disable mmx. It has some check for this, but > it seems it doesn't work properly. I'll check it later. Much better IMO to edit jhbuildrc as you've done than to try patching and regenerating configure. I would just leave a FIXME comment with a link to this bug so that it can be reverted in the future. But really, I wonder why you're using clang for the dependencies. I understand using it for WebKit itself, but for an entire jhbuild moduleset of dependencies outside our control, I bet something will always be broken.
(In reply to comment #4) > (In reply to comment #3) > > (In reply to comment #2) > > > If this is a problem only for clang, would it be possible to detect it > > > somehow from jhbuildrc and add the autogen command line argument there? > > > > I don't know if it is possible, but maybe it is easier to patch > > configure.ac to disable mmx. It has some check for this, but > > it seems it doesn't work properly. I'll check it later. > > Much better IMO to edit jhbuildrc as you've done than to try patching and > regenerating configure. I would just leave a FIXME comment with a link to > this bug so that it can be reverted in the future. It can be much better if I know how to do it. But now I have no idea if it is possible to detect the compiler by jhbuild. And I don't have any time to investigate jhbuild internals in the near future. > But really, I wonder why you're using clang for the dependencies. I > understand using it for WebKit itself, but for an entire jhbuild moduleset > of dependencies outside our control, I bet something will always be broken. Because I don't want to switch compiler settings between jhbuild and build-webkit. The final goal is to make everything work with GCC ang clang too.
(In reply to comment #5) > It can be much better if I know how to do it. But now I have no idea > if it is possible to detect the compiler by jhbuild. And I don't have > any time to investigate jhbuild internals in the near future. You can do it in the .jhbuildrc by checking for CC in the environment, but that's just as yucky as a patch. You can't do it from the moduleset file. Anyway, I would simply disable the thing unconditionally, just as you've done, so it's the same regardless of which compiler is used. I like your current patch as-is; I would just also add FIXME comments above those changes.
Comment on attachment 265859 [details] Patch Ok, let's follow the easiest approach, then.
Created attachment 266199 [details] Patch Patch for landing with additional FIXME comment.
Created attachment 266200 [details] Patch Patch for landing with additional FIXME comment.
Committed r192780: <http://trac.webkit.org/changeset/192780>