http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50035
Bug #: 50035
Summary: [4.7 Regression] FAIL: g++.dg/tree-prof/partition2.C
execution, -Os -fprofile-use at revision 176696 on
x86_64-apple-darwin10
Classification: Unclassified
Product: gcc
Version: 4.7.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: middle-end
AssignedTo: unassig...@gcc.gnu.org
ReportedBy: domi...@lps.ens.fr
CC: hjl.to...@gmail.com, r...@gcc.gnu.org
Host: x86_64-apple-darwin10
Target: x86_64-apple-darwin10
Build: x86_64-apple-darwin10
Created attachment 24969
--> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24969
Working assembly
Since revision 176696, the execution of g++.dg/tree-prof/partition2.C with -m64
-Os -fprofile-use gives a segmentation fault on x86_64-apple-darwin10. The test
succeeds on powerpc-apple-darwin9.
The difference between the working and faulty assembly files is
--- partition2.s-176695p1 2011-08-10 11:45:45.000000000 +0200
+++ partition2.s-176696p1 2011-08-10 12:13:42.000000000 +0200
@@ -36,21 +36,18 @@ LEHB2:
call ___cxa_throw
LEHE2:
LVL1:
-L6:
- jmp L4
-L7:
- jmp L5
- .section __TEXT,__text_cold,regular,pure_instructions
-___cold_sect_of_main:
L4:
movq %rax, %rbp
+ jmp L2
+ .section __TEXT,__text_cold,regular,pure_instructions
+___cold_sect_of_main:
+L2:
movq %rbx, %rdi
call ___cxa_free_exception
LVL2:
movq %rbp, %rax
jmp L3
L5:
- jmp L3
L3:
LM5:
movq %rax, %rdi
@@ -89,14 +86,14 @@ LLSDA0:
.long L$set$2
.set L$set$3,LEHE1-LEHB1
.long L$set$3
- .set L$set$4,L6-LFB0
+ .set L$set$4,L4-LFB0
.long L$set$4
.byte 0x3
.set L$set$5,LEHB2-LFB0
.long L$set$5
.set L$set$6,LEHE2-LEHB2
.long L$set$6
- .set L$set$7,L7-LFB0
+ .set L$set$7,L5-LFB0
.long L$set$7
.byte 0x1
.byte 0x1
@@ -308,7 +305,7 @@ Ldebug_info0:
.long L$set$32
.byte 0x8
.byte 0x1
- .ascii "GNU C++ 4.7.0 20110723 (experimental) [trunk revision 176695p1]\0"
+ .ascii "GNU C++ 4.7.0 20110723 (experimental) [trunk revision 176696p1]\0"
.byte 0x4
.ascii "partition2.C\0"
.ascii "/Users/dominiq/Documents/Fortran/g95bench/win/f90/bug\0"
Note that on i686-pc-linux-gnu the compilation fails at revision 176696 (see
http://gcc.gnu.org/ml/gcc-testresults/2011-07/msg02767.html) while the test
succeeds at revision 176688 (see
http://gcc.gnu.org/ml/gcc-testresults/2011-07/msg02743.html ). The error is
Executing on host: /home/regress/tbox/native/build/gcc/testsuite/g++/../../g++
-B/home/regress/tbox/native/build/gcc/testsuite/g++/../../
/home/regress/tbox/svn-gcc/gcc/testsuite/g++.dg/tree-prof/partition2.C
-nostdinc++
-I/home/regress/tbox/native/build/i686-pc-linux-gnu/libstdc++-v3/include/i686-pc-linux-gnu
-I/home/regress/tbox/native/build/i686-pc-linux-gnu/libstdc++-v3/include
-I/home/regress/tbox/svn-gcc/libstdc++-v3/libsupc++
-I/home/regress/tbox/svn-gcc/libstdc++-v3/include/backward
-I/home/regress/tbox/svn-gcc/libstdc++-v3/testsuite/util -fmessage-length=0
-Os -fnon-call-exceptions -freorder-blocks-and-partition -fprofile-use
-L/home/regress/tbox/native/build/i686-pc-linux-gnu/./libstdc++-v3/src/.libs
-B/home/regress/tbox/native/build/i686-pc-linux-gnu/./libstdc++-v3/src/.libs
-L/home/regress/tbox/native/build/i686-pc-linux-gnu/./libstdc++-v3/src/.libs
-lm -o /home/regress/tbox/native/build/gcc/testsuite/g++/partition2.x62
(timeout = 300)
/tmp/ccUB3whI.s: Assembler messages:
/tmp/ccUB3whI.s:95: Error: invalid sections for operation on `.L5' and `.LFB0'
compiler exited with status 1
output is:
/tmp/ccUB3whI.s: Assembler messages:
/tmp/ccUB3whI.s:95: Error: invalid sections for operation on `.L5' and `.LFB0'
FAIL: g++.dg/tree-prof/partition2.C compilation, -Os -fprofile-use
UNRESOLVED: g++.dg/tree-prof/partition2.C execution, -Os -fprofile-use
The same kind of failure can also be seen on x86_64-unknown-linux-gnu and
i686-pc-linux-gnu (see
http://gcc.gnu.org/ml/gcc-testresults/2011-07/msg02770.html and
http://gcc.gnu.org/ml/gcc-testresults/2011-07/msg02765.html).
I have no idea if the failures on linux are due to the same origin than on
darwin and no easy access to regstrapping on linux. Feel free to open another
pr for linux!-)
This pr may also be related/duplicate of pr49972.
