[llvm] r200695 - Don't use -ffunction-sections if -fno-function-sections is not supported in the compiler.

[Alexey Samsonov]

I've submitted r200745, which is probably a better fix.


On Tue, Feb 4, 2014 at 11:47 AM, Evgeniy Stepanov <eugeni.stepanov at gmail.com
> wrote:

> Well, it's not that easy. Nothing is easy with cmake. I'll try some
> more today, and either fix or revert the change.
>
>
> On Tue, Feb 4, 2014 at 1:49 AM, Justin Bogner <mail at justinbogner.com>
> wrote:
> > Evgeniy Stepanov <eugeni.stepanov at gmail.com> writes:
> >> Don't use -ffunction-sections if -fno-function-sections is not
> >> supported in the compiler.
> >>
> >> This will disable -ffunction-sections in older versions of Clang where
> it
> >> breaks build of sanitizer runtime library.
> >>
> >> Modified:
> >>     llvm/trunk/cmake/modules/HandleLLVMOptions.cmake
> >>
> >> Modified: llvm/trunk/cmake/modules/HandleLLVMOptions.cmake
> >> URL:
> >>
> http://llvm.org/viewvc/llvm-project/llvm/trunk/cmake/modules/HandleLLVMOptions.cmake?rev=200695&r1=200694&r2=200695&view=diff==============================================================================
> >> --- llvm/trunk/cmake/modules/HandleLLVMOptions.cmake (original)
> >> +++ llvm/trunk/cmake/modules/HandleLLVMOptions.cmake Mon Feb  3
> 07:57:09 2014
> >> @@ -350,13 +350,23 @@ if (UNIX AND
> >>    append("-fcolor-diagnostics" CMAKE_C_FLAGS CMAKE_CXX_FLAGS)
> >>  endif()
> >>
> >> +# Clang prior to 3.5 ignored -fno-function-sections.
> >> +# It's pretty hard to test directly, so we rely on the version number.
> >> +if( ("${CMAKE_CXX_COMPILER_ID}" MATCHES "Clang") AND
> (CMAKE_CXX_COMPILER_VERSION VERSION_LESS 3.5) )
> >> +  set(LLVM_COMPILER_HAS_BROKEN_FNO_FUNCTION_SECTIONS ON)
> >> +endif()
> >> +
> >
> > Please don't do it this way. Relying on version numbers to check for a
> > feature is the wrong thing to do. It doesn't account for the 5000
> > revisions that are called 3.5 and don't support this flag and just
> > generally isn't nice to packagers.
> >
> > It also wouldn't be very scalable if other compilers needed to be
> > handled as well, though that doesn't seem to be the case in this
> > particular example.
> >
> > Can't we check if the compiler supports the flag directly?
> >
> >>  # Add flags for add_dead_strip().
> >>  # FIXME: With MSVS, consider compiling with /Gy and linking with
> /OPT:REF?
> >>  # But MinSizeRel seems to add that automatically, so maybe disable
> these
> >>  # flags instead if LLVM_NO_DEAD_STRIP is set.
> >>  if(NOT CYGWIN AND NOT WIN32)
> >>    if(NOT ${CMAKE_SYSTEM_NAME} MATCHES "Darwin")
> >> -    append("-ffunction-sections -fdata-sections" CMAKE_C_FLAGS
> CMAKE_CXX_FLAGS)
> >> +    # Don't add -ffunction-section if it can be disabled with
> -fno-function-sections.
> >> +    # Doing so will break sanitizers.
> >> +    if (NOT LLVM_COMPILER_HAS_BROKEN_FNO_FUNCTION_SECTIONS)
> >> +      append("-ffunction-sections -fdata-sections" CMAKE_C_FLAGS
> CMAKE_CXX_FLAGS)
> >> +    endif()
> >>    endif()
> >>  endif()
> >>
> >>
> >>
> >> _______________________________________________
> >> llvm-commits mailing list
> >> llvm-commits at cs.uiuc.edu
> >> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>



-- 
Alexey Samsonov, MSK
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20140204/8acaf5eb/attachment.html>