[zorg] r270539 - Enable modules selfhosting builds with libstdc++, too.
Vassil Vassilev via llvm-commits
llvm-commits at lists.llvm.org
Sat Jun 11 11:38:58 PDT 2016
On 09/06/16 01:21, Richard Smith wrote:
> On Thu, May 26, 2016 at 2:21 AM, Vassil Vassilev
> <v.g.vassilev at gmail.com <mailto:v.g.vassilev at gmail.com>> wrote:
>
> On 26/05/16 00:42, Richard Smith wrote:
>> Seems really weird to have a modules flag to getClangBuildFactory
>> that doesn't actually turn on -DLLVM_ENABLE_MODULES. What's the
>> reason for removing this:
>>
>> - if modules:
>> - extra_args = ['-DLLVM_ENABLE_MODULES=1']
>>
>> ?
> I wanted to move all extra_args into stage2_extra_configure_args.
> I can revert this part. However it seems that the only use of the
> modules flag is go add a "prebuild" step
> "compile.llvm.stage2.intrinsics_gen". Do we need this, on my
> machine it works just fine without?
>
>
> Yes, we need this. A -j build will otherwise *sometimes* fail,
> depending on whether the steps that consume the IR module happen to
> start before or after intrinsics.gen is generated by tablegen.
>
> We could fix this by splitting the LLVM IR module up into the pieces
> that use intrinsics.gen and the pieces that don't; I don't think we're
> actually missing build dependencies here, but the relevant modules are
> too coarse.
AFAICT, intrinsics.gen generates Attributes.h and Intrinsics.h. I split
LLVM_IR headers that #include either Attributes.h or Intrinsics.h into
separate submodule (http://reviews.llvm.org/D21258). Would that make sense?
>
>> On Wed, May 25, 2016 at 10:17 AM, Vassil Vassilev
>> <v.g.vassilev at gmail.com <mailto:v.g.vassilev at gmail.com>> wrote:
>>
>> Done in r270725. Thanks for reviewing!
>> -- Vassil
>>
>> On 25/05/16 19:00, Galina Kistanova wrote:
>>> Looks ok, please commit it.
>>>
>>> Thanks
>>>
>>> Galina
>>>
>>>
>>>
>>> On Wed, May 25, 2016 at 5:19 AM, Vassil Vassilev
>>> <v.g.vassilev at gmail.com <mailto:v.g.vassilev at gmail.com>> wrote:
>>>
>>> Hi Galina,
>>> I am attaching a patch that I believe fixes the issue.
>>> Could you review it please (I don't know how to test it).
>>> Many thanks,
>>> Vassil
>>>
>>> On 25/05/16 11:36, Vassil Vassilev wrote:
>>>
>>> This broke our modules slaves. I am working on a
>>> fix. Sorry for the inconvenience!
>>> -- Vassil
>>> On 24/05/16 11:23, Vassil Vassilev via llvm-commits
>>> wrote:
>>>
>>> Author: vvassilev
>>> Date: Tue May 24 04:22:58 2016
>>> New Revision: 270539
>>>
>>> URL:
>>> http://llvm.org/viewvc/llvm-project?rev=270539&view=rev
>>> Log:
>>> Enable modules selfhosting builds with
>>> libstdc++, too.
>>>
>>> If a modulemap is not present (in the case of
>>> libstdc++), we will end up with
>>> duplicate entities in the modules. This will
>>> help us stress test the module
>>> merging and discover potential bugs.
>>>
>>> Modified:
>>> zorg/trunk/zorg/buildbot/builders/ClangBuilder.py
>>>
>>> Modified:
>>> zorg/trunk/zorg/buildbot/builders/ClangBuilder.py
>>> URL:
>>> http://llvm.org/viewvc/llvm-project/zorg/trunk/zorg/buildbot/builders/ClangBuilder.py?rev=270539&r1=270538&r2=270539&view=diff
>>> ==============================================================================
>>>
>>> ---
>>> zorg/trunk/zorg/buildbot/builders/ClangBuilder.py
>>> (original)
>>> +++
>>> zorg/trunk/zorg/buildbot/builders/ClangBuilder.py
>>> Tue May 24 04:22:58 2016
>>> @@ -281,8 +281,6 @@ def getClangBuildFactory(
>>> cxx_flags = ''
>>> extra_args = []
>>> if modules:
>>> - # Modules requires libc++ for now (we
>>> don't have a module map for libstdc++ yet).
>>> - cxx_flags += '-stdlib=libc++'
>>> extra_args = ['-DLLVM_ENABLE_MODULES=1']
>>> f.addStep(ShellCommand(name='cmake',
>>>
>>>
>>> _______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at lists.llvm.org
>>> <mailto:llvm-commits at lists.llvm.org>
>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>>
>>>
>>>
>>>
>>>
>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20160611/64c2bee5/attachment.html>
More information about the llvm-commits
mailing list