https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62247
--- Comment #5 from dave.anglin at bell dot net --- On 2015-02-07, at 10:49 AM, jakub at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62247 > > --- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> --- > With cross-compiler I get the same anon3.s (no .weak occurrences in the > assembly) as in 4.9. As neither the test nor dg-require-weak seems to have > changed, I guess the important questions are: > 1) can you compare 4.9.2 and 5.0.0 generated assembly? Attached assembler output from 4.9.3 20150207. The assembly output from 5.0.0 is wierd. It has a few nop's which make up body of function: virtual void Heya::A::f() ... (insn 2 11 3 (set (mem/f/c:SI (reg/f:SI 28 %r28 [98]) [0 this+0 S4 A32]) (reg:SI 26 %r26 [ this ])) /test/gnu/gcc/gcc/gcc/testsuite/g++.dg/abi/an on3.C:14 40 {*pa.md:2204} (nil)) (note 3 2 8 NOTE_INSN_FUNCTION_BEG)(insn 8 3 18 (const_int 0 [0]) /test/gnu/gcc/gcc/gcc/testsuite/g++.dg/abi/anon3. C:14 213 {nop} (nil)) (note 18 8 19 NOTE_INSN_EPILOGUE_BEG) > 2) does the test in 4.9.2 PASS or is it UNSUPPORTED? The test passes in 4.9.2. > 3) has something changed in auto-host.h, such as previously HAVE_GAS_WEAK and > now no longer true or something similar? Both 4.9 and 5.0 have: /* Define if your assembler supports .weak. */ #ifndef USED_FOR_TARGET #define HAVE_GAS_WEAK 1 #endif /* Define if your assembler supports .weakref. */ #ifndef USED_FOR_TARGET #define HAVE_GAS_WEAKREF 1 #endif > 4) if there has been any change on the compiler side, can you bisect when did > that happen? Will try to narrow this down. Dave -- John David Anglin dave.ang...@bell.net
