[cfe-commits] r165988 - in /cfe/trunk: include/clang/Basic/ObjCRuntime.h lib/CodeGen/CGObjC.cpp lib/Driver/ToolChains.cpp lib/Driver/Tools.cpp lib/Sema/SemaExpr.cpp runtime/compiler-rt/Makefile test/CodeGenObjC/optimized-setter-ios-device.m test/

[Nico Weber]

With the verbose patch I see all commands, but all ranlib invocations
are run with `2> /dev/null`, so I still don't get error output. But it
looks like you worked around that somehow already :-)

If disabling the arm bits on <= 10.6 fixes things, that sounds good to
me. Let me know if you want me to test any patches.

On Wed, Oct 17, 2012 at 9:36 AM, Jean-Daniel Dupas
<devlists at shadowlab.org> wrote:
>
> Le 17 oct. 2012 à 18:18, Jean-Daniel Dupas <devlists at shadowlab.org> a écrit :
>
>>
>> Le 17 oct. 2012 à 17:55, Nico Weber <thakis at chromium.org> a écrit :
>>
>>> On Wed, Oct 17, 2012 at 1:03 AM, Jean-Daniel Dupas
>>> <devlists at shadowlab.org> wrote:
>>>>
>>>> Le 17 oct. 2012 à 08:10, Nico Weber <thakis at chromium.org> a écrit :
>>>>
>>>>> On Tue, Oct 16, 2012 at 10:33 PM, Bob Wilson <bob.wilson at apple.com> wrote:
>>>>>>
>>>>>> On Oct 16, 2012, at 9:25 PM, Bob Wilson <bob.wilson at apple.com> wrote:
>>>>>>
>>>>>>>
>>>>>>> On Oct 16, 2012, at 9:18 PM, Nico Weber <thakis at chromium.org> wrote:
>>>>>>>> It's also pretty disrupting to land a CL that breaks the build and
>>>>>>>> then ignore the bug report for it for 2 weeks.
>>>>>>>
>>>>>>> Yeah, sorry about the slow response.
>>>>>>>
>>>>>>>>
>>>>>>>> Repro steps are:
>>>>>>>> 1. Get a 10.6 box
>>>>>>>
>>>>>>> I don't have a 10.6 box.  Certainly someone at Apple must have one around, but it will take me longer to find one that I can use.  I was hoping for a quick fix.
>>>>>>
>>>>>> PR14102 looks like a different issue than PR14013, where the problem was that the armv7 slice was dropped from the cc_kext_ios5 config, leaving it with no valid architectures at all.  In your case, it is going ahead and building armv7 for both cc_kext and cc_kext_ios5 but failing intermittently, apparently when creating the archive.
>>>>>>
>>>>>> You say that you have a 10.6 box.  What version of Xcode do you have installed?
>>>>>
>>>>> 3.2.6
>>>>>
>>>>>> Did you also install the command-line tools package?
>>>>>
>>>>> All versions of Xcode that run on 10.6 precede the command-line tools
>>>>> package thingy as far as I know. (3.2.6 does for sure.)
>>>>
>>>> The command line package tool does exist in pre Xcode 4 versions, but it is exposed as an optional package that you can check/uncheck in the installer.
>>>> I can't remember if is was check by default though.
>>>
>>> I do have libtool and co in my path (and didn't do anything to make
>>> that happen). Also, if I revert the iOS 6 changes locally, things
>>> build fine, so I doubt that's related.
>>
>>
>> If you apply the attached patch to clang, and build using "make VERBOSE=1" it should print all compilation commands (including runtime libraries compilation).
>>
>> Maybe it can help to nail down the source of the problem.
>>
>> I have a 10.6 machine and will try to reproduce the bug too.
>
>
> I found an issue too. It crashes on the "ar" invocation used to build the armv7 compiler-rt library for cc_kext with the error:
>
> /usr/bin/ranlib: object: /Users/jddupas/Projects/llvm/simplebuild/tools/clang/runtime/compiler-rt/clang_darwin/cc_kext/armv7/libcompiler_rt.a(absvdi2.o) malformed object (unknown load command 1)
> ar: internal ranlib command failed
>
>
> -- Jean-Daniel
>
>
>
>