svn commit: r367883 - in head: Mk lang/gcc

[Adam Weinberger]

On 11 Sep, 2014, at 18:38, Bryan Drewery <bdrewery at FreeBSD.org> wrote:

> On 9/10/2014 4:32 PM, Gerald Pfeifer wrote:
>> On Wed, 10 Sep 2014, Bryan Drewery wrote:
>>> Thanks for writing up a summary of changes. Very useful. I learned 
>>> about AddressSanitizer being supported.
>> 
>> You are very welcome!  (I haven't tested AddressSanitizer myself, 
>> but think I recall a positive report by a user.  If it does not 
>> work properly we might look into lang/gcc49 or lang/gcc5.)
> 
> Actually it fails to link. Both clang and gcc complain about missing
> libasan symbols.
> 
> I find no providers of libasan in any of the gcc, llvm or clang packages.
> 
> It appears it is a submodule of both gcc and clang. So it likely just
> need to be hooked up in the build for gcc.
> 
> Simple test case:
> 
> # echo 'int main() {return 0;}'|g++5 -Wall -fsanitize=address -o
> /dev/null -x c -
> /usr/local/bin/ld: cannot find -lasan
> collect2: error: ld returned 1 exit status

From lang/gcc/work/gcc-4.8.3/libsanitizer/configure.tgt:
case "${target}" in
  x86_64-*-linux* | i?86-*-linux*)
        if test x$ac_cv_sizeof_void_p = x8; then
                TSAN_SUPPORTED=yes
        fi
        ;;
  powerpc*le-*-linux*)
        UNSUPPORTED=1
        ;;
  powerpc*-*-linux*)
        ;;
  sparc*-*-linux*)
        ;;
  x86_64-*-darwin[1]* | i?86-*-darwin[1]*)
        TSAN_SUPPORTED=no
        ;;
  *)
        UNSUPPORTED=1
        ;;
esac

In configure output:
checking for libsanitizer support… no

# Adam


-- 
Adam Weinberger
adamw at adamw.org
http://www.adamw.org


> 
> 
>> 
>> 
>> On a related note, when trying to commit the PORTREVISION bump 
>> related to this commit, I ended up with:
>> 
>>  svn: E165001: Commit failed (details follow):
>>  svn: E165001: Commit blocked by pre-commit hook (exit code 1) with output:
>>  Do not commit ports with NOPORTDOCS or NOPORTEXAMPLES.
>>  The port must be converted to proper OPTIONS.  See
>>  http://www.freebsd.org/doc/en_US.ISO8859-1/books/porters-handbook/makefile-options.html
>> 
>> Not super helpful since it did not indicate which of the 1000+ ports 
>> actually failed this test. ;-)  I could bypass this using Approved:
>> portmgr (implicit).
>> 
>> Gerald
>> 
> 
> 
> -- 
> Regards,
> Bryan Drewery