[test-suite] r210788 - disable the -lsupc++ specification for powerpc.

[Will Schmidt]

On Thu, 2014-06-12 at 15:14 +0000, Will Schmidt wrote:
> Author: willschm
> Date: Thu Jun 12 10:14:00 2014
> New Revision: 210788
> 
> URL: http://llvm.org/viewvc/llvm-project?rev=210788&view=rev
> Log:
> disable the -lsupc++ specification for powerpc.


As seen on the buildbot here:
http://lab.llvm.org:8011/builders/clang-ppc64-elf-linux/builds/8750/steps/lnt.nightly-test/logs/test.log

specifically this snippet:
/home/buildbots/clangslave1/clang-ppc64-1/test-suite/RunSafely.sh --show-errors -t "/home/buildbots/clangslave1/clang-ppc64-1/tests/nt/build/sample-0/tools/timeit" 500 /dev/null Output/utils.llvm.o.compile \
  /home/buildbots/clangslave1/clang-ppc64-1/llvm.install.1/bin/clang -DCOUNTBITS16 -DLASTBIT16 -DCOUNTMOVES_TABLE -DTWO_STAGE_GENERATION -DHASHCODEBITS=23 -I/home/buildbots/clangslave1/clang-ppc64-1/tests/nt/build/sample-0/MultiSource/Applications/obsequi -I/home/buildbots/clangslave1/clang-ppc64-1/test-suite/MultiSource/Applications/obsequi -I/home/buildbots/clangslave1/clang-ppc64-1/test-suite/include -I../../../include -D_GNU_SOURCE -D__STDC_LIMIT_MACROS -DNDEBUG  -O3 -mcpu=native -ffp-contract=off -fomit-frame-pointer -c /home/buildbots/clangslave1/clang-ppc64-1/test-suite/MultiSource/Applications/obsequi/utils.c -o Output/utils.llvm.o
/home/buildbots/clangslave1/clang-ppc64-1/llvm.install.1/bin/clang++ -o Output/Obsequi.simple Output/display.llvm.o Output/does_x_win.llvm.o Output/hash.llvm.o Output/init.llvm.o Output/move_gen.llvm.o Output/move_sort.llvm.o Output/negamax.llvm.o Output/obsequi.llvm.o Output/position_values.llvm.o Output/tables.llvm.o Output/toggle_move.llvm.o Output/traits.llvm.o Output/utils.llvm.o -lsupc++  -mcpu=native -ffp-contract=off -fomit-frame-pointer
/bin/ld: cannot find -lsupc++
clang: error: linker command failed with exit code 1 (use -v to see invocation)

The build fails to find libsupc++.so.  
Having scoured the (fedora) yum repositories for that particular .so,
and asking a few questions of the fedora guys, the response that best
seems to fit what I'm seeing is 
" afaik libsupc++ is a subset of libstdc++ for static linking, so there's no .so of that "

A manual test shows that the "-lsupc++" is not necessary for a
successful link of obsequi.   A quick hack (as committed here) has
allowed the ppc64 LNT bot to successfully pass the obsequi test.  (still
running at the time of this email, but it's definitely past the point of
previous failure).
http://lab.llvm.org:8011/builders/clang-ppc64-elf-linux/builds/8753/steps/lnt.nightly-test/logs/test.log


I've limited the hack to powerpc arch to limit affecting the other
arches...  I suspect it should not be necessary for the other arches,
but then do not understand why the "-lsupc++" does not trigger a fail
elsewhere...

Thoughts? 

Thanks, 
-Will



> Modified:
>     test-suite/trunk/MultiSource/Applications/obsequi/Makefile
> 
> Modified: test-suite/trunk/MultiSource/Applications/obsequi/Makefile
> URL: http://llvm.org/viewvc/llvm-project/test-suite/trunk/MultiSource/Applications/obsequi/Makefile?rev=210788&r1=210787&r2=210788&view=diff
> ==============================================================================
> --- test-suite/trunk/MultiSource/Applications/obsequi/Makefile (original)
> +++ test-suite/trunk/MultiSource/Applications/obsequi/Makefile Thu Jun 12 10:14:00 2014
> @@ -9,7 +9,10 @@ else
>  CPPFLAGS += -DHASHCODEBITS=23
>  endif
> 
> +ifneq ($(ARCH),PowerPC)
>  LDFLAGS = -lsupc++
>  LIBS += -lsupc++
> +endif
> +
>  STDIN_FILENAME=$(PROJ_SRC_DIR)/input
>  include ../../Makefile.multisrc
> 
> 
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
>