[Bug lto/85574] [9 Regression] LTO bootstapped binaries differ

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85574

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2018-05-01 00:00:00         |2018-12-20

--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
Re-confirmed on trunk.  prev-gcc/cc1 vs gcc/cc1 has

-  57add1:      e8 de e3 17 00          callq  6f91b4
<_ZL12bmp_iter_setP15bitmap_iteratorPj.part.0.lto_priv.0>
+  57add1:      e8 e0 e3 17 00          callq  6f91b6
<_ZL12bmp_iter_setP15bitmap_iteratorPj.part.0.lto_priv.0>
...

@@ -22292,50 +22292,50 @@
   590038:      be fc 03 00 00          mov    $0x3fc,%esi
   59003d:      bf 60 35 74 01          mov    $0x1743560,%edi
   590042:      e8 8f e9 fe ff          callq  57e9d6 <_Z11fancy_abortPKciS0_>
-  590047:      ba cc 2d c7 01          mov    $0x1c72dcc,%edx
-  59004c:      be 1a 05 00 00          mov    $0x51a,%esi
-  590051:      bf 60 35 74 01          mov    $0x1743560,%edi
-  590056:      e8 7b e9 fe ff          callq  57e9d6 <_Z11fancy_abortPKciS0_>
-  59005b:      45 31 c9                xor    %r9d,%r9d
...
+  590047:      41 b8 c5 28 74 01       mov    $0x17428c5,%r8d
+  59004d:      b9 73 01 00 00          mov    $0x173,%ecx
+  590052:      ba 28 be 5b 01          mov    $0x15bbe28,%edx
+  590057:      4c 89 ef                mov    %r13,%rdi
+  59005a:      be 0f 00 00 00          mov    $0xf,%esi
+  59005f:      e8 bf 46 07 00          callq  604723
<_Z33tree_contains_struct_
check_failedPK9tree_node24tree_node_structure_enumPKciS4_>

which looks like different inlining decisions happening?!

I'd have to re-run with IPA dumps but there's definitely sth suspicious
happening.

I'll see if I can reproduce the same builds at least (so two times prev-gcc/cc1
are equal).