http://sourceware.org/bugzilla/show_bug.cgi?id=14342
Jan Hubicka <hubicka at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |hubicka at gcc dot gnu.org
Severity|normal |critical
--- Comment #1 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
This problem now reprduce with mainline GCC indirect call profiling test
gcc.dg/tree-prof/crossmodule-indircall-1.c:
evans:/abuild/jh/trunk-3/build-inst11-check/gcc/:[2]# ./xgcc -B ./ -O3 -flto
-fprofile-generate
../../gcc/testsuite/gcc.dg/tree-prof/crossmodule-indircall-1*.c --save-temps
/abuild/jh/trunk-install/x86_64-unknown-linux-gnu/bin/ld:
crossmodule-indircall-1.o: previous definition here
/abuild/jh/trunk-install/x86_64-unknown-linux-gnu/bin/ld: error:
./libgcov.a(_gcov_indirect_call_profiler.o): symbol
'__gcov_indirect_call_callee' used as both __thread and non-__thread
/abuild/jh/trunk-install/x86_64-unknown-linux-gnu/bin/ld:
crossmodule-indircall-1.o: previous definition here
while it links just fine with GNU-ld or with gold and -fno-lto.
The symbol is used as TLS symbol. non-LTO uses:
crossmodule-indircall-1.s: movq
__gcov_indirect_call_callee@gottpoff(%rip), %r12
crossmodule-indircall-1a.s: movq
__gcov_indirect_call_callee@gottpoff(%rip), %rax
LTO uses:
ccsUXr3s.ltrans0.s: movq __gcov_indirect_call_callee@gottpoff(%rip),
%rbp
and libgcc definition is:
leaq __gcov_indirect_call_callee@tlsgd(%rip), %rdi
.globl __gcov_indirect_call_callee
.type __gcov_indirect_call_callee, @object
.size __gcov_indirect_call_callee, 8
__gcov_indirect_call_callee:
.long __gcov_indirect_call_callee@dtpoff, 0
.string "__gcov_indirect_call_callee"
I am not an expert on TLS, but it seems all right to me, since all goes through
tls and gottpoff or dtpoff.
I am not really happy about leaving -fprofile-generate broken with -flto on
mainline. Is there a workaround? I really need this TLS variable...
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
bug-binutils mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/bug-binutils
