[cfe-dev] clang on darwin broken since r212689

[Jack Howarth]

The actual problem code is…

install(
  FILES ${files} ${CMAKE_CURRENT_BINARY_DIR}/arm_neon.h
  PERMISSIONS OWNER_READ OWNER_WRITE GROUP_READ WORLD_READ
  DESTINATION lib${LLVM_LIBDIR_SUFFIX}/clang/${CLANG_VERSION}/include)

in lib/Headers/CMakeLists.txt. From
http://www.cmake.org/Wiki/CMake:Install_Commands,
it would appear that the entries in FILES are reduced to the base file name
and any prefixed
path is stripped off. A separate files2 list would be required for
_stdint.h and _types.h as well
as an explicit install with the DESTINATION
lib${LLVM_LIBDIR_SUFFIX}/clang/${CLANG_VERSION}/include/machine.



On Fri, Jul 18, 2014 at 4:29 PM, Richard Smith <richard at metafoo.co.uk>
wrote:

> FYI, I've requested a revert.
>
>
> On Fri, Jul 18, 2014 at 1:02 PM, Jack Howarth <
> howarth.mailing.lists at gmail.com> wrote:
>
>> Ismail,
>>      Unfortunately that doesn't address the actual bug. The problem code
>> is in
>>
>> set(out_files)foreach( f ${files} )
>>   set( src ${CMAKE_CURRENT_SOURCE_DIR}/${f} )
>>   set( dst ${output_dir}/${f} )
>>   add_custom_command(OUTPUT ${dst}
>>     DEPENDS ${src}
>>     COMMAND ${CMAKE_COMMAND} -E copy_if_different ${src} ${dst}
>>     COMMENT "Copying clang's ${f}...")
>>   list(APPEND out_files ${dst})endforeach( f )
>>
>>
>> of lib/Headers/CMakeLists.txt which fails to create the machine
>> subdirectory during 'make install'.
>> The entries for machine/_stdint.h and machine/_types.h in files are
>> reduced to just _stdint.h and
>> _types.h when the those files entries are copied. This is the bug that
>> need to be fixed and not
>> hacked around.
>>          Jack
>>
>>
>> On Fri, Jul 18, 2014 at 1:32 PM, Ismail Pazarbasi <
>> ismail.pazarbasi at gmail.com> wrote:
>>
>>> On Fri, Jul 18, 2014 at 5:48 PM, Jack Howarth
>>> <howarth.mailing.lists at gmail.com> wrote:
>>> >     I haven't seen this discussed outside of my bug report…
>>> >
>>> > http://llvm.org/bugs/show_bug.cgi?id=20326
>>> >
>>> > and the later comments on…
>>> >
>>> > http://reviews.llvm.org/D3908
>>> >
>>> > but the commit of r212689…
>>> >
>>> > https://www.mail-archive.com/cfe-commits@cs.uiuc.edu/msg100897.html
>>> >
>>> > has broken the darwin targets for clang via blockage of the proper use
>>> of
>>> > the
>>> > system _stdint.h and _types.h headers by the newly added clang
>>> versions.
>>> > This should be a release blocker for branching 3.5 until it is
>>> resolved, no?
>>> > Deleting the installed
>>> /sw/opt/llvm-3.5.0/lib/clang/3.5.0/include/_stdint.h
>>> > and
>>> > /sw/opt/llvm-3.5.0/lib/clang/3.5.0/include/_types.h files eliminates
>>> the
>>> > breakage
>>> > so perhaps the best solution is not to have those files installed at
>>> all
>>> > outside
>>> > of the freebsd target.
>>> > Jack
>>> >
>>> > _______________________________________________
>>> > cfe-dev mailing list
>>> > cfe-dev at cs.uiuc.edu
>>> > http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
>>> >
>>>
>>> Great timing, I was planning to write about this just a few minutes
>>> ago. I didn't investigate which commit broke it. I have prepared a
>>> tiny patch that I applied locally. basically, on top of the header
>>> file:
>>>
>>> #if defined(__APPLE__)
>>> #include_next<_types.h>
>>> #endif
>>>
>>> I guess we can address this issue either with something like my patch
>>> or by skipping installation on darwin.
>>>
>>> _______________________________________________
>>> cfe-dev mailing list
>>> cfe-dev at cs.uiuc.edu
>>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
>>>
>>
>>
>> _______________________________________________
>> cfe-dev mailing list
>> cfe-dev at cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-dev
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20140718/0169a438/attachment.html>