[cfe-dev] Problems compiling clang 3.7.0 with gcc 4.9

[Richard Smith via cfe-dev]

On Wed, Oct 7, 2015 at 5:24 PM, Craig Rodrigues via cfe-dev <
cfe-dev at lists.llvm.org> wrote:

> On Wed, Oct 7, 2015 at 1:23 PM, Richard Smith <richard at metafoo.co.uk>
> wrote:
>
>> On Wed, Oct 7, 2015 at 4:47 AM, Craig Rodrigues <rodrigc at freebsd.org>
>> wrote:
>>
>>> Hi,
>>>
>>> I recently tried to compile clang 3.7.0 with gcc 4.9
>>> under FreeBSD.  I reproduced the problem with gcc 5.0.
>>>
>>> The full build error is here:
>>> https://jenkins.freebsd.org/job/FreeBSD_HEAD_amd64_gcc4.9/617/console
>>>
>>> The error of interest is:
>>>
>>>
>>> --- Module.o ---
>>> /usr/local/bin/x86_64-portbld-freebsd10.1-g++
>>> -isystem
>>> /builds/FreeBSD_HEAD_amd64_gcc4.9/obj/builds/FreeBSD_HEAD_amd64_gcc4.9/tmp/usr/include
>>> -L/builds/FreeBSD_HEAD_amd64_gcc4.9/obj/builds/FreeBSD_HEAD_amd64_gcc4.9/tmp/usr/lib
>>> --sysroot=/builds/FreeBSD_HEAD_amd64_gcc4.9/obj/builds/FreeBSD_HEAD_amd64_gcc4.9/tmp
>>> -B/usr/local/x86_64-freebsd/bin/
>>> -I/builds/FreeBSD_HEAD_amd64_gcc4.9/obj/builds/FreeBSD_HEAD_amd64_gcc4.9/tmp/usr/include/c++/v1
>>> -std=gnu++11
>>> -L/builds/FreeBSD_HEAD_amd64_gcc4.9/obj/builds/FreeBSD_HEAD_amd64_gcc4.9/tmp/../lib/libc++
>>> --sysroot=/builds/FreeBSD_HEAD_amd64_gcc4.9/obj/builds/FreeBSD_HEAD_amd64_gcc4.9/tmp
>>> -B/usr/local/x86_64-freebsd/bin/  -O2 -pipe
>>> -I/builds/FreeBSD_HEAD_amd64_gcc4.9/lib/clang/libclangbasic/../../../contrib/llvm/include
>>> -I/builds/FreeBSD_HEAD_amd64_gcc4.9/lib/clang/libclangbasic/../../../contrib/llvm/tools/clang/include
>>> -I/builds/FreeBSD_HEAD_amd64_gcc4.9/lib/clang/libclangbasic/../../../contrib/llvm/tools/clang/lib/Basic
>>> -I.
>>> -I/builds/FreeBSD_HEAD_amd64_gcc4.9/lib/clang/libclangbasic/../../../contrib/llvm/../../lib/clang/include
>>> -DLLVM_ON_UNIX -DLLVM_ON_FREEBSD -D__STDC_LIMIT_MACROS
>>> -D__STDC_CONSTANT_MACROS -DCLANG_ENABLE_ARCMT
>>> -DCLANG_ENABLE_STATIC_ANALYZER -fno-strict-aliasing
>>> -DLLVM_DEFAULT_TARGET_TRIPLE=\"x86_64-unknown-freebsd11.0\"
>>> -DLLVM_HOST_TRIPLE=\"x86_64-unknown-freebsd11.0\" -DDEFAULT_SYSROOT=\"\"
>>> -fstack-protector-strong  -std=c++11 -fno-exceptions -fno-rtti  -c
>>> /builds/FreeBSD_HEAD_amd64_gcc4.9/lib/clang/libclangbasic/../../../contrib/llvm/tools/clang/lib/Basic/Module.cpp
>>> -o Module.o
>>> In file included from
>>> /builds/FreeBSD_HEAD_amd64_gcc4.9/lib/clang/libclangbasic/../../../contrib/llvm/tools/clang/include/clang/Basic/SourceLocation.h:22:0,
>>>                  from
>>>                  /builds/FreeBSD_HEAD_amd64_gcc4.9/lib/clang/libclangbasic/../../../contrib/llvm/tools/clang/include/clang/Basic/Module.h:18,
>>>                  from
>>>                  /builds/FreeBSD_HEAD_amd64_gcc4.9/lib/clang/libclangbasic/../../../contrib/llvm/tools/clang/lib/Basic/Module.cpp:15:
>>> /builds/FreeBSD_HEAD_amd64_gcc4.9/obj/builds/FreeBSD_HEAD_amd64_gcc4.9/tmp/usr/include/c++/v1/functional:1322:17:
>>> error: '_Rp std::__1::__function::__base<_Rp(_ArgTypes
>>> ...)>::operator()(_ArgTypes&& ...) [with _Rp = void; _ArgTypes =
>>> {clang::VisibleModuleSet::setVisible(clang::Module*, clang::SourceLocation,
>>> clang::VisibleModuleSet::VisibleCallback,
>>> clang::VisibleModuleSet::ConflictCallback)::Visiting}]', declared
>>> using local type 'clang::VisibleModuleSet::setVisible(clang::Module*,
>>> clang::SourceLocation, clang::VisibleModuleSet::VisibleCallback,
>>> clang::VisibleModuleSet::ConflictCallback)::Visiting', is used but
>>> never defined [-fpermissive]
>>>      virtual _Rp operator()(_ArgTypes&& ...) = 0;
>>>                  ^
>>>
>>> This looks like a GCC bug. I'm not sure why it imagines that this pure
>> virtual function is used; did it produce any note indicating that?
>>
>
> There was no note to indicate why GCC thought that this pure virtual
> function is used.
> The only thing I could think of was maybe there was something in the llvm
> implementation of
> std::function that caused this to happen.
>
>
>
>> In any case, given that the link succeeds if you move the struct
>> definition outside of the function (so it's not able to detect the problem
>> during compilation any more), this function is very likely not actually
>> used, and thus the diagnostic here is wrong.
>>
>> One more thing that's worth checking: if you build with -fpermissive,
>> does the compile and link succeed?
>>
>>
> I added -fpermissive to the command-line, and the compile and link
> succeeded.
>
> If this looks like a GCC bug, I can submit something to them.
>

Yes, this seems likely to be a GCC bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20151007/1e21fc01/attachment.html>