[Bug java/41260] New: major regressions in libjava testsuite between 4.4.1 and gcc trunk

[Bug java/41260] New: major regressions in libjava testsuite between 4.4.1 and gcc trunk

[howarth at nitro dot med dot uc dot edu]

We have accumulated a large number of regressions in the libjava testsuite on
the x86_64-apple-darwin10 target in current gcc trunk compared to gcc 4.4.1...

http://gcc.gnu.org/ml/gcc-testresults/2009-09/msg00294.html
http://gcc.gnu.org/ml/gcc-testresults/2009-07/msg01866.html

Currently we have 133 failures on r151394 compared to only 14 for gcc 4.4.1.
These failures go back to at least 20090721 (137 failures). We will need to
perform a  manual regression hunt to try to bracket the revisions where these
were introduced after gcc 4.4 branched. Also we need to check the
x86_64-apple-darwin9 target for these regressions.


-- 
           Summary: major regressions in libjava testsuite between 4.4.1 and
                    gcc trunk
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: java
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: howarth at nitro dot med dot uc dot edu
 GCC build triplet: x86_64-apple-darwin10
  GCC host triplet: x86_64-apple-darwin10
GCC target triplet: x86_64-apple-darwin10


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] major regressions in libjava testsuite between 4.4.1 and gcc trunk

[dominiq at lps dot ens dot fr]

------- Comment #1 from dominiq at lps dot ens dot fr  2009-09-04 11:44 -------
On i686-apple-darwin9 my latest regtest (r151159) shows:

                === libjava tests ===


Running target unix

                === libjava Summary for unix ===

# of expected passes            2574

Running target unix/-m64
FAIL: PR16923 run
FAIL: events output

                === libjava Summary for unix/-m64 ===

# of expected passes            2572
# of unexpected failures        2
# of untested testcases         1

                === libjava Summary ===

# of expected passes            5146
# of unexpected failures        2
# of untested testcases         1


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] major regressions in libjava testsuite between 4.4.1 and gcc trunk

[howarth at nitro dot med dot uc dot edu]

------- Comment #2 from howarth at nitro dot med dot uc dot edu  2009-09-04 12:50 -------
Can you try testing x86_64-apple-darwin9 so I an focus on a regression hunt for
x86_64-apple-darwin10?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] major regressions in libjava testsuite between 4.4.1 and gcc trunk

[howarth at nitro dot med dot uc dot edu]

------- Comment #3 from howarth at nitro dot med dot uc dot edu  2009-09-04 12:53 -------
Dominique,
    Also if you are bothering to run the test suite on i686-apple-darwin9
periodically, you might as well shoot the results over to the gcc-testresults
mailing list (since Apple has never set up a i686-apple-darwin9 regress
server).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] major regressions in libjava testsuite between 4.4.1 and gcc trunk

[howarth at nitro dot med dot uc dot edu]

------- Comment #4 from howarth at nitro dot med dot uc dot edu  2009-09-05 01:24 -------
First round of regression hunting shows that r146781 (4/25/09) passes the
libjava like gcc 4.4.1 but r148013 (5/31/09) shows the additional 120+ libjava
failures.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] major regressions in libjava testsuite between 4.4.1 and gcc trunk

[howarth at nitro dot med dot uc dot edu]

------- Comment #5 from howarth at nitro dot med dot uc dot edu  2009-09-05 07:14 -------
Second round of regression hunting shows..

r147974 without regressions

r147995 with regressions


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] major regressions in libjava testsuite between 4.4.1 and gcc trunk

[howarth at nitro dot med dot uc dot edu]

------- Comment #6 from howarth at nitro dot med dot uc dot edu  2009-09-05 08:08 -------
r147982 without regressions.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] major regressions in libjava testsuite between 4.4.1 and gcc trunk

[howarth at nitro dot med dot uc dot edu]

------- Comment #7 from howarth at nitro dot med dot uc dot edu  2009-09-05 08:57 -------
r147989 without regressions. The regressions start with r147995.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] major regressions in libjava testsuite between 4.4.1 and gcc trunk caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #8 from howarth at nitro dot med dot uc dot edu  2009-09-05 08:58 -------
Author: rth
Date: Sat May 30 00:33:46 2009
New Revision: 147995

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=147995
Log:
        * cfgcleanup.c (try_crossjump_to_edge): Only skip past
        NOTE_INSN_BASIC_BLOCK.
        * cfglayout.c (duplicate_insn_chain): Copy epilogue insn marks.
        Duplicate NOTE_INSN_EPILOGUE_BEG notes.
        * cfgrtl.c (can_delete_note_p): Allow NOTE_INSN_EPILOGUE_BEG
        to be deleted.
        * dwarf2out.c (struct cfa_loc): Change indirect field to bitfield,
        add in_use field.
        (add_cfi): Disable check redefining cfa away from drap.
        (lookup_cfa_1): Add remember argument; handle remember/restore.
        (lookup_cfa): Pass remember argument.
        (cfa_remember): New.
        (compute_barrier_args_size_1): Remove sibcall check.
        (dwarf2out_frame_debug_def_cfa): New.
        (dwarf2out_frame_debug_adjust_cfa): New.
        (dwarf2out_frame_debug_cfa_offset): New.
        (dwarf2out_frame_debug_cfa_register): New.
        (dwarf2out_frame_debug_cfa_restore): New.
        (dwarf2out_frame_debug): Handle REG_CFA_* notes.
        (dwarf2out_begin_epilogue): New.
        (dwarf2out_frame_debug_restore_state): New.
        (dw_cfi_oprnd1_desc): Handle DW_CFA_remember_state,
        DW_CFA_restore_state.
        (output_cfi_directive): Likewise.
        (convert_cfa_to_fb_loc_list): Likewise.
        (dw_cfi_oprnd1_desc): Handle DW_CFA_restore.
        * dwarf2out.h: Update.
        * emit-rtl.c (try_split): Don't split RTX_FRAME_RELATED_P.
        (copy_insn_1): Early out for null.
        * final.c (final_scan_insn): Call dwarf2out_begin_epilogue
        and dwarf2out_frame_debug_restore_state.
        * function.c (prologue, epilogue, sibcall_epilogue): Remove.
        (prologue_insn_hash, epilogue_insn_hash): New.
        (free_after_compilation): Adjust freeing accordingly.
        (record_insns): Create hash table if needed; push insns into
        hash instead of array.
        (maybe_copy_epilogue_insn): New.
        (contains): Search hash table instead of array.
        (sibcall_epilogue_contains): Remove.
        (thread_prologue_and_epilogue_insns): Split eh_return insns
        and mark them as epilogues.
        (reposition_prologue_and_epilogue_notes): Rewrite epilogue
        scanning in terms of basic blocks.
        * insn-notes.def (CFA_RESTORE_STATE): New.
        * jump.c (returnjump_p_1): Accept EH_RETURN.
        (eh_returnjump_p_1, eh_returnjump_p): New.
        * reg-notes.def (CFA_DEF_CFA, CFA_ADJUST_CFA, CFA_OFFSET,
        CFA_REGISTER, CFA_RESTORE): New.
        * rtl.def (EH_RETURN): New.
        * rtl.h (eh_returnjump_p, maybe_copy_epilogue_insn): Declare.

        * config/bfin/bfin.md (UNSPEC_VOLATILE_EH_RETURN): Remove.
        (eh_return_internal): Use eh_return rtx; split w/ epilogue.

        * config/i386/i386.c (gen_push): Update cfa state.
        (pro_epilogue_adjust_stack): Add set_cfa argument.  When true,
        add a CFA_ADJUST_CFA note.
        (ix86_dwarf_handle_frame_unspec): Remove.
        (ix86_expand_prologue): Update cfa state.
        (ix86_emit_restore_reg_using_pop): New.
        (ix86_emit_restore_regs_using_pop): New.
        (ix86_emit_leave): New.
        (ix86_emit_restore_regs_using_mov): Add CFA_RESTORE notes.
        (ix86_expand_epilogue): Add notes for unwinding the epilogue.
        * config/i386/i386.h (struct machine_cfa_state): New.
        (ix86_cfa_state): New.
        * config/i386/i386.md (UNSPEC_EH_RETURN): Remove.
        (eh_return_internal): Merge from eh_return_<mode>,
        use eh_return rtx, split w/ epilogue.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] [4.5 Regression] major regressions in libjava testsuite between 4.4.1 and gcc trunk caused by r147995

[rguenth at gcc dot gnu dot org]

-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|major regressions in libjava|[4.5 Regression] major
                   |testsuite between 4.4.1 and |regressions in libjava
                   |gcc trunk caused by r147995 |testsuite between 4.4.1 and
                   |                            |gcc trunk caused by r147995
   Target Milestone|---                         |4.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] [4.5 Regression] major regressions in libjava testsuite between 4.4.1 and gcc trunk caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #9 from howarth at nitro dot med dot uc dot edu  2009-09-05 13:35 -------
Created an attachment (id=18505)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18505&action=view)
testsuite results for r147989 on x86_64-apple-darwin10


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] [4.5 Regression] major regressions in libjava testsuite between 4.4.1 and gcc trunk caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #10 from howarth at nitro dot med dot uc dot edu  2009-09-05 14:02 -------
Still running make check on r147995 but that revision appears to have caused
massive regressions in testsuite which aren't present in currently in gcc
trunk. Will look for any other test cases outside of libjava that are still
failing on x86_64-apple-darwin10 and started failing at r147995.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] [4.5 Regression] major regressions in libjava testsuite between 4.4.1 and gcc trunk caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #11 from howarth at nitro dot med dot uc dot edu  2009-09-05 14:27 -------
I just noticed PR 40304 which fixed regressions stack unwind from r147995. Will
move testing up to that revision to try to reduce the noise when looking for
residual regressions in current gcc trunk that may have been introduced at
r147995 (outside of libjava).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug java/41260] [4.5 Regression] major regressions in libjava testsuite between 4.4.1 and gcc trunk caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #12 from howarth at nitro dot med dot uc dot edu  2009-09-05 15:25 -------
In r147989, we had no regressions in x86_64-apple-darwin10 in the g++
testsuite. However, in r148013 (which is after the fix for PR 40304), we have
the following g++ testsuite failures which still exist in gcc trunk on
x86_64-apple-darwin10...

FAIL: g++.dg/cpp0x/variadic73.C execution test
FAIL: g++.dg/eh/delayslot1.C execution test
FAIL: g++.dg/eh/ia64-2.C execution test
FAIL: g++.dg/eh/simd-1.C execution test
FAIL: g++.dg/eh/simd-2.C execution test
FAIL: g++.dg/eh/synth2.C execution test
FAIL: g++.dg/ext/cleanup-8.C execution test
FAIL: g++.dg/ext/cleanup-9.C execution test
FAIL: g++.dg/init/array16.C execution test
FAIL: g++.dg/init/array5.C execution test
FAIL: g++.dg/init/ref9.C execution test
FAIL: g++.dg/opt/eh2.C execution test
FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O0  execution test
FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O1  execution test
FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O2  execution test
FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O3 -fomit-frame-pointer 
execution test
FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O3 -g  execution test
FAIL: g++.dg/torture/stackalign/throw-1.C  -O0  execution test
FAIL: g++.dg/torture/stackalign/throw-1.C  -O1  execution test
FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O0  execution test
FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O1  execution test
FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O2  execution test
FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O3 -fomit-frame-pointer 
execution test
FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O3 -g  execution test
FAIL: g++.dg/torture/stackalign/throw-1.C  -O0  execution test
FAIL: g++.dg/torture/stackalign/throw-1.C  -O1  execution test
FAIL: g++.old-deja/g++.abi/cxa_vec.C execution test
FAIL: g++.old-deja/g++.eh/badalloc1.C execution test
FAIL: g++.old-deja/g++.eh/rethrow3.C execution test
FAIL: g++.old-deja/g++.eh/vbase2.C execution test
FAIL: g++.old-deja/g++.martin/new1.C execution test
FAIL: g++.old-deja/g++.mike/eh2.C execution test
FAIL: g++.old-deja/g++.mike/eh29.C execution test
FAIL: g++.old-deja/g++.mike/eh41.C execution test
FAIL: g++.old-deja/g++.mike/eh44.C execution test
FAIL: g++.old-deja/g++.oliva/delete3.C execution test
FAIL: g++.old-deja/g++.other/array1.C execution test
FAIL: g++.old-deja/g++.other/init7.C execution test
FAIL: g++.old-deja/g++.other/singleton.C execution test
FAIL: g++.old-deja/g++.robertl/eb50.C execution test
FAIL: g++.old-deja/g++.robertl/eh990323-5.C execution test

Any idea of which of these would present the easy opportunity for debugging the
problem?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions in libjava testsuite between 4.4.1 and gcc trunk caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #13 from howarth at nitro dot med dot uc dot edu  2009-09-05 15:32 -------
We also picked up failures in the gcc testsuite between r147989 and r1478013
of...

FAIL: gcc.dg/cleanup-12.c execution test
FAIL: gcc.dg/cleanup-8.c execution test

in case these are directly related to this regression.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on x86_64-apple-darwin10 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #14 from howarth at nitro dot med dot uc dot edu  2009-09-05 15:58 -------
Looking at the assembly output from the testcase...

FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O0  execution test

the diff between that generated in r147989 and r1478013 is...

---
/sw/src/fink.build/gcc45-4.4.999-20090529/darwin_objdir/gcc/testsuite/g++/eh-vararg-2.s.r147989
    2009-09-05 11:53:50.000000000 -0400
+++
/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/gcc/testsuite/g++/eh-vararg-2.s.r148013
    2009-09-05 11:48:40.000000000 -0400
@@ -26,19 +26,20 @@
        movq    24(%rsp), %rax
        movl    (%rax), %eax
        leave
+LCFI4:
        ret
 LFE0:
 .globl __Z5checkPvi
 __Z5checkPvi:
 LFB1:
        pushq   %rbp
-LCFI4:
-       movq    %rsp, %rbp
 LCFI5:
-       andq    $-32, %rsp
+       movq    %rsp, %rbp
 LCFI6:
-       subq    $32, %rsp
+       andq    $-32, %rsp
 LCFI7:
+       subq    $32, %rsp
+LCFI8:
        movq    %rdi, 24(%rsp)
        movl    %esi, 20(%rsp)
        movl    20(%rsp), %eax
@@ -51,6 +52,7 @@
        call    _abort
 L3:
        leave
+LCFI9:
        ret
 LFE1:
 .globl _global
@@ -66,13 +68,13 @@
 __Z3barPci:
 LFB2:
        pushq   %rbp
-LCFI8:
+LCFI10:
        movq    %rsp, %rbp
-LCFI9:
+LCFI11:
        andq    $-32, %rsp
-LCFI10:
+LCFI12:
        subq    $32, %rsp
-LCFI11:
+LCFI13:
        movq    %rdi, 24(%rsp)
        movl    %esi, 20(%rsp)
        movl    20(%rsp), %eax
@@ -82,6 +84,7 @@
        movq    %rax, %rdi
        call    _strncpy
        leave
+LCFI14:
        ret
 LFE2:
        .section __TEXT,__textcoal_nt,coalesced,pure_instructions
@@ -91,11 +94,12 @@
 __ZN4BaseC2Ev:
 LFB5:
        pushq   %rbp
-LCFI12:
+LCFI15:
        movq    %rsp, %rbp
-LCFI13:
+LCFI16:
        movq    %rdi, -8(%rbp)
        leave
+LCFI17:
        ret
 LFE5:
        .align 1
@@ -104,13 +108,13 @@
 __ZN1AC1Ev:
 LFB8:
        pushq   %rbp
-LCFI14:
+LCFI18:
        movq    %rsp, %rbp
-LCFI15:
+LCFI19:
        andq    $-32, %rsp
-LCFI16:
+LCFI20:
        subq    $32, %rsp
-LCFI17:
+LCFI21:
        movq    %rdi, 24(%rsp)
        movq    24(%rsp), %rax
        movq    %rax, %rdi
@@ -120,6 +124,7 @@
        movq    24(%rsp), %rax
        movq    %rdx, (%rax)
        leave
+LCFI22:
        ret
 LFE8:
        .text
@@ -127,21 +132,21 @@
 __Z4testP13__va_list_tag:
 LFB9:
        leaq    8(%rsp), %r10
-LCFI18:
+LCFI23:
        andq    $-64, %rsp
-LCFI19:
+LCFI24:
        pushq   -8(%r10)
-LCFI20:
+LCFI25:
        pushq   %rbp
-LCFI21:
+LCFI26:
        movq    %rsp, %rbp
-LCFI22:
+LCFI27:
        pushq   %r10
-LCFI23:
+LCFI28:
        pushq   %rbx
-LCFI24:
+LCFI29:
        subq    $160, %rsp
-LCFI25:
+LCFI30:
        movq    %rdi, -152(%rbp)
        movq    -152(%rbp), %rax
        movl    (%rax), %eax
@@ -306,13 +311,13 @@
 __Z3fooPKcz:
 LFB13:
        pushq   %rbp
-LCFI26:
+LCFI31:
        movq    %rsp, %rbp
-LCFI27:
+LCFI32:
        andq    $-32, %rsp
-LCFI28:
+LCFI33:
        subq    $256, %rsp
-LCFI29:
+LCFI34:
        movq    %rsi, 72(%rsp)
        movq    %rdx, 80(%rsp)
        movq    %rcx, 88(%rsp)
@@ -348,6 +353,7 @@
        movq    %rax, %rdi
        call    __Z4testP13__va_list_tag
        leave
+LCFI35:
        ret
 LFE13:
        .const
@@ -358,13 +364,13 @@
 _main:
 LFB14:
        pushq   %rbp
-LCFI30:
+LCFI36:
        movq    %rsp, %rbp
-LCFI31:
+LCFI37:
        andq    $-32, %rsp
-LCFI32:
+LCFI38:
        subq    $32, %rsp
-LCFI33:
+LCFI39:
        movsd   LC1(%rip), %xmm0
        movl    $5, %esi
        leaq    LC2(%rip), %rdi
@@ -388,6 +394,7 @@
 L27:
        movl    $0, %eax
        leave
+LCFI40:
        ret
 LFE14:
        .section __DATA,__gcc_except_tab
@@ -523,155 +530,190 @@
        .long L$set$14
        .byte   0xd
        .byte   0x6
+       .byte   0x4
+       .set L$set$15,LCFI4-LCFI1
+       .long L$set$15
+       .byte   0xc6
+       .byte   0xc
+       .byte   0x7
+       .byte   0x8
        .align 3
 LEFDE1:
 .globl __Z5checkPvi.eh
 __Z5checkPvi.eh:
 LSFDE3:
-       .set L$set$15,LEFDE3-LASFDE3
-       .long L$set$15
+       .set L$set$16,LEFDE3-LASFDE3
+       .long L$set$16
 LASFDE3:
        .long   LASFDE3-EH_frame1
        .quad   LFB1-.
-       .set L$set$16,LFE1-LFB1
-       .quad L$set$16
+       .set L$set$17,LFE1-LFB1
+       .quad L$set$17
        .byte   0x8
        .quad   0x0
        .byte   0x4
-       .set L$set$17,LCFI4-LFB1
-       .long L$set$17
+       .set L$set$18,LCFI5-LFB1
+       .long L$set$18
        .byte   0xe
        .byte   0x10
        .byte   0x11
        .byte   0x6
        .byte   0x2
        .byte   0x4
-       .set L$set$18,LCFI5-LCFI4
-       .long L$set$18
+       .set L$set$19,LCFI6-LCFI5
+       .long L$set$19
        .byte   0xd
        .byte   0x6
+       .byte   0x4
+       .set L$set$20,LCFI9-LCFI6
+       .long L$set$20
+       .byte   0xc6
+       .byte   0xc
+       .byte   0x7
+       .byte   0x8
        .align 3
 LEFDE3:
 .globl __Z3barPci.eh
 __Z3barPci.eh:
 LSFDE5:
-       .set L$set$19,LEFDE5-LASFDE5
-       .long L$set$19
+       .set L$set$21,LEFDE5-LASFDE5
+       .long L$set$21
 LASFDE5:
        .long   LASFDE5-EH_frame1
        .quad   LFB2-.
-       .set L$set$20,LFE2-LFB2
-       .quad L$set$20
+       .set L$set$22,LFE2-LFB2
+       .quad L$set$22
        .byte   0x8
        .quad   0x0
        .byte   0x4
-       .set L$set$21,LCFI8-LFB2
-       .long L$set$21
+       .set L$set$23,LCFI10-LFB2
+       .long L$set$23
        .byte   0xe
        .byte   0x10
        .byte   0x11
        .byte   0x6
        .byte   0x2
        .byte   0x4
-       .set L$set$22,LCFI9-LCFI8
-       .long L$set$22
+       .set L$set$24,LCFI11-LCFI10
+       .long L$set$24
        .byte   0xd
        .byte   0x6
+       .byte   0x4
+       .set L$set$25,LCFI14-LCFI11
+       .long L$set$25
+       .byte   0xc6
+       .byte   0xc
+       .byte   0x7
+       .byte   0x8
        .align 3
 LEFDE5:
 .globl __ZN4BaseC2Ev.eh
        .weak_definition __ZN4BaseC2Ev.eh
 __ZN4BaseC2Ev.eh:
 LSFDE7:
-       .set L$set$23,LEFDE7-LASFDE7
-       .long L$set$23
+       .set L$set$26,LEFDE7-LASFDE7
+       .long L$set$26
 LASFDE7:
        .long   LASFDE7-EH_frame1
        .quad   LFB5-.
-       .set L$set$24,LFE5-LFB5
-       .quad L$set$24
+       .set L$set$27,LFE5-LFB5
+       .quad L$set$27
        .byte   0x8
        .quad   0x0
        .byte   0x4
-       .set L$set$25,LCFI12-LFB5
-       .long L$set$25
+       .set L$set$28,LCFI15-LFB5
+       .long L$set$28
        .byte   0xe
        .byte   0x10
        .byte   0x11
        .byte   0x6
        .byte   0x2
        .byte   0x4
-       .set L$set$26,LCFI13-LCFI12
-       .long L$set$26
+       .set L$set$29,LCFI16-LCFI15
+       .long L$set$29
        .byte   0xd
        .byte   0x6
+       .byte   0x4
+       .set L$set$30,LCFI17-LCFI16
+       .long L$set$30
+       .byte   0xc6
+       .byte   0xc
+       .byte   0x7
+       .byte   0x8
        .align 3
 LEFDE7:
 .globl __ZN1AC1Ev.eh
        .weak_definition __ZN1AC1Ev.eh
 __ZN1AC1Ev.eh:
 LSFDE9:
-       .set L$set$27,LEFDE9-LASFDE9
-       .long L$set$27
+       .set L$set$31,LEFDE9-LASFDE9
+       .long L$set$31
 LASFDE9:
        .long   LASFDE9-EH_frame1
        .quad   LFB8-.
-       .set L$set$28,LFE8-LFB8
-       .quad L$set$28
+       .set L$set$32,LFE8-LFB8
+       .quad L$set$32
        .byte   0x8
        .quad   0x0
        .byte   0x4
-       .set L$set$29,LCFI14-LFB8
-       .long L$set$29
+       .set L$set$33,LCFI18-LFB8
+       .long L$set$33
        .byte   0xe
        .byte   0x10
        .byte   0x11
        .byte   0x6
        .byte   0x2
        .byte   0x4
-       .set L$set$30,LCFI15-LCFI14
-       .long L$set$30
+       .set L$set$34,LCFI19-LCFI18
+       .long L$set$34
        .byte   0xd
        .byte   0x6
+       .byte   0x4
+       .set L$set$35,LCFI22-LCFI19
+       .long L$set$35
+       .byte   0xc6
+       .byte   0xc
+       .byte   0x7
+       .byte   0x8
        .align 3
 LEFDE9:
 .globl __Z4testP13__va_list_tag.eh
 __Z4testP13__va_list_tag.eh:
 LSFDE11:
-       .set L$set$31,LEFDE11-LASFDE11
-       .long L$set$31
+       .set L$set$36,LEFDE11-LASFDE11
+       .long L$set$36
 LASFDE11:
        .long   LASFDE11-EH_frame1
        .quad   LFB9-.
-       .set L$set$32,LFE9-LFB9
-       .quad L$set$32
+       .set L$set$37,LFE9-LFB9
+       .quad L$set$37
        .byte   0x8
        .quad   LLSDA9-.
        .byte   0x4
-       .set L$set$33,LCFI18-LFB9
-       .long L$set$33
+       .set L$set$38,LCFI23-LFB9
+       .long L$set$38
        .byte   0xc
        .byte   0xa
        .byte   0x0
        .byte   0x4
-       .set L$set$34,LCFI21-LCFI18
-       .long L$set$34
+       .set L$set$39,LCFI26-LCFI23
+       .long L$set$39
        .byte   0x10
        .byte   0x6
        .byte   0x2
        .byte   0x76
        .byte   0x0
        .byte   0x4
-       .set L$set$35,LCFI23-LCFI21
-       .long L$set$35
+       .set L$set$40,LCFI28-LCFI26
+       .long L$set$40
        .byte   0xf
        .byte   0x3
        .byte   0x76
        .byte   0x78
        .byte   0x6
        .byte   0x4
-       .set L$set$36,LCFI25-LCFI23
-       .long L$set$36
+       .set L$set$41,LCFI30-LCFI28
+       .long L$set$41
        .byte   0x10
        .byte   0x3
        .byte   0x2
@@ -682,55 +724,69 @@
 .globl __Z3fooPKcz.eh
 __Z3fooPKcz.eh:
 LSFDE13:
-       .set L$set$37,LEFDE13-LASFDE13
-       .long L$set$37
+       .set L$set$42,LEFDE13-LASFDE13
+       .long L$set$42
 LASFDE13:
        .long   LASFDE13-EH_frame1
        .quad   LFB13-.
-       .set L$set$38,LFE13-LFB13
-       .quad L$set$38
+       .set L$set$43,LFE13-LFB13
+       .quad L$set$43
        .byte   0x8
        .quad   0x0
        .byte   0x4
-       .set L$set$39,LCFI26-LFB13
-       .long L$set$39
+       .set L$set$44,LCFI31-LFB13
+       .long L$set$44
        .byte   0xe
        .byte   0x10
        .byte   0x11
        .byte   0x6
        .byte   0x2
        .byte   0x4
-       .set L$set$40,LCFI27-LCFI26
-       .long L$set$40
+       .set L$set$45,LCFI32-LCFI31
+       .long L$set$45
        .byte   0xd
        .byte   0x6
+       .byte   0x4
+       .set L$set$46,LCFI35-LCFI32
+       .long L$set$46
+       .byte   0xc6
+       .byte   0xc
+       .byte   0x7
+       .byte   0x8
        .align 3
 LEFDE13:
 .globl _main.eh
 _main.eh:
 LSFDE15:
-       .set L$set$41,LEFDE15-LASFDE15
-       .long L$set$41
+       .set L$set$47,LEFDE15-LASFDE15
+       .long L$set$47
 LASFDE15:
        .long   LASFDE15-EH_frame1
        .quad   LFB14-.
-       .set L$set$42,LFE14-LFB14
-       .quad L$set$42
+       .set L$set$48,LFE14-LFB14
+       .quad L$set$48
        .byte   0x8
        .quad   LLSDA14-.
        .byte   0x4
-       .set L$set$43,LCFI30-LFB14
-       .long L$set$43
+       .set L$set$49,LCFI36-LFB14
+       .long L$set$49
        .byte   0xe
        .byte   0x10
        .byte   0x11
        .byte   0x6
        .byte   0x2
        .byte   0x4
-       .set L$set$44,LCFI31-LCFI30
-       .long L$set$44
+       .set L$set$50,LCFI37-LCFI36
+       .long L$set$50
        .byte   0xd
        .byte   0x6
+       .byte   0x4
+       .set L$set$51,LCFI40-LCFI37
+       .long L$set$51
+       .byte   0xc6
+       .byte   0xc
+       .byte   0x7
+       .byte   0x8
        .align 3
 LEFDE15:
        .constructor


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on x86_64-apple-darwin10 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #15 from howarth at nitro dot med dot uc dot edu  2009-09-05 16:01 -------
Created an attachment (id=18510)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18510&action=view)
g++.dg/torture/stackalign/eh-vararg-2.C preprocessed source on
x86_64-apple-darwin10 at r147989


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on x86_64-apple-darwin10 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #16 from howarth at nitro dot med dot uc dot edu  2009-09-05 16:02 -------
Created an attachment (id=18511)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18511&action=view)
g++.dg/torture/stackalign/eh-vararg-2.C assembly on x86_64-apple-darwin10 at
r147989


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on x86_64-apple-darwin10 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #17 from howarth at nitro dot med dot uc dot edu  2009-09-05 16:03 -------
Created an attachment (id=18512)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18512&action=view)
g++.dg/torture/stackalign/eh-vararg-2.C assembly on x86_64-apple-darwin10 at
r148013


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on x86_64-apple-darwin10 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #18 from howarth at nitro dot med dot uc dot edu  2009-09-05 16:05 -------
Attached assembly from failing...

FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O0  execution test

introduced by r147995 was produced for r147989 and r1478013 using the
commands...

/sw/src/fink.build/gcc45-4.4.999-20090529/darwin_objdir/gcc/testsuite/g++/../../g++
-B/sw/src/fink.build/gcc45-4.4.999-20090529/darwin_objdir/gcc/testsuite/g++/../../
/sw/src/fink.build/gcc45-4.4.999-20090529/gcc-4.5-20090529/gcc/testsuite/g++.dg/torture/stackalign/eh-vararg-2.C
--save-temps -nostdinc++
-I/sw/src/fink.build/gcc45-4.4.999-20090529/darwin_objdir/x86_64-apple-darwin10/libstdc++-v3/include/x86_64-apple-darwin10
-I/sw/src/fink.build/gcc45-4.4.999-20090529/darwin_objdir/x86_64-apple-darwin10/libstdc++-v3/include
-I/sw/src/fink.build/gcc45-4.4.999-20090529/gcc-4.5-20090529/libstdc++-v3/libsupc++
-I/sw/src/fink.build/gcc45-4.4.999-20090529/gcc-4.5-20090529/libstdc++-v3/include/backward
-I/sw/src/fink.build/gcc45-4.4.999-20090529/gcc-4.5-20090529/libstdc++-v3/testsuite/util
-fmessage-length=0 -O0 -mstackrealign -mpreferred-stack-boundary=5
-L/sw/src/fink.build/gcc45-4.4.999-20090529/darwin_objdir/x86_64-apple-darwin10/./libstdc++-v3/src/.libs
-L/sw/src/fink.build/gcc45-4.4.999-20090529/darwin_objdir/x86_64-apple-darwin10/./libstdc++-v3/src/.libs
-L/sw/src/fink.build/gcc45-4.4.999-20090529/darwin_objdir/x86_64-apple-darwin10/./libiberty
-multiply_defined suppress -lm -o ./eh-vararg-2.exe

and

/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/gcc/testsuite/g++/../../g++
-B/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/gcc/testsuite/g++/../../
/sw/src/fink.build/gcc45-4.4.999-20090531/gcc-4.5-20090531/gcc/testsuite/g++.dg/torture/stackalign/eh-vararg-2.C
--save-temps -nostdinc++
-I/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/libstdc++-v3/include/x86_64-apple-darwin10
-I/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/libstdc++-v3/include
-I/sw/src/fink.build/gcc45-4.4.999-20090531/gcc-4.5-20090531/libstdc++-v3/libsupc++
-I/sw/src/fink.build/gcc45-4.4.999-20090531/gcc-4.5-20090531/libstdc++-v3/include/backward
-I/sw/src/fink.build/gcc45-4.4.999-20090531/gcc-4.5-20090531/libstdc++-v3/testsuite/util
-fmessage-length=0 -O0 -mstackrealign -mpreferred-stack-boundary=5
-L/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/./libstdc++-v3/src/.libs
-L/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/./libstdc++-v3/src/.libs
-L/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/./libiberty
-multiply_defined suppress -lm -o ./eh-vararg-2.exe

respectively.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on x86_64-apple-darwin10 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #19 from howarth at nitro dot med dot uc dot edu  2009-09-05 16:11 -------
Richard,
    Please let me know if there is a better test case than
g++.dg/torture/stackalign/eh-vararg-2.C, out of the regressions caused by
r147995, listed in Comment 12 or 13 and I'll provide the assembly output from
under r147989 and r1478013 for that case as well.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on x86_64-apple-darwin10 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #20 from howarth at nitro dot med dot uc dot edu  2009-09-05 16:23 -------
In case it matters, the eh-vararg-2.exe test case in r148013 fails with the
message...

terminate called after throwing an instance of 'A'


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on x86_64-apple-darwin10 caused by r147995

[jakub at gcc dot gnu dot org]

------- Comment #21 from jakub at gcc dot gnu dot org  2009-09-05 19:21 -------
What unwinder are you using?  The freshly built libgcc_s, or some outdated
darwin system libgcc_s?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on x86_64-apple-darwin10 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #22 from howarth at nitro dot med dot uc dot edu  2009-09-05 19:34 -------
I am using the from the gcc build. The test case binary is built as...

otool -L eh-vararg-2.exe 
eh-vararg-2.exe:
        /usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current
version 124.1.1)
        /sw/lib/gcc4.5/lib/libstdc++.6.dylib (compatibility version 7.0.0,
current version 7.12.0)
        /sw/lib/gcc4.5/lib/libgcc_s.1.dylib (compatibility version 1.0.0,
current version 1.0.0)

but it should be using the shared libraries from with in the build tree during
the make check...

Executing on host:
/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/gcc/testsuite/g++/../../g++
-B/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/gcc/testsuite/g++/../../
/sw/src/fink.build/gcc45-4.4.999-20090531/gcc-4.5-20090531/gcc/testsuite/g++.dg/torture/stackalign/eh-vararg-2.C
 -nostdinc++
-I/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/libstdc++-v3/include/x86_64-apple-darwin10
-I/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/libstdc++-v3/include
-I/sw/src/fink.build/gcc45-4.4.999-20090531/gcc-4.5-20090531/libstdc++-v3/libsupc++
-I/sw/src/fink.build/gcc45-4.4.999-20090531/gcc-4.5-20090531/libstdc++-v3/include/backward
-I/sw/src/fink.build/gcc45-4.4.999-20090531/gcc-4.5-20090531/libstdc++-v3/testsuite/util
-fmessage-length=0  -O0  -mstackrealign -mpreferred-stack-boundary=5   
-L/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/./libstdc++-v3/src/.libs

-L/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/./libstdc++-v3/src/.libs
-L/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/./libiberty
 -multiply_defined suppress -lm   -o ./eh-vararg-2.exe    (timeout = 300)
PASS: g++.dg/torture/stackalign/eh-vararg-2.C  -O0  (test for excess errors)
Setting LD_LIBRARY_PATH to
.:/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/./libstdc++-v3/src/.libs:/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/./libstdc++-v3/src/.libs:/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/gcc:.:/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/./libstdc++-v3/src/.libs:/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/x86_64-apple-darwin10/./libstdc++-v3/src/.libs:/sw/src/fink.build/gcc45-4.4.999-20090531/darwin_objdir/gcc
terminate called after throwing an instance of 'A'
FAIL: g++.dg/torture/stackalign/eh-vararg-2.C  -O0  execution test


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on x86_64-apple-darwin10 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #23 from howarth at nitro dot med dot uc dot edu  2009-09-05 22:05 -------
These failures in the g++ and libjava testsuites do not occur for
x86_64-apple-darwin9 with current gcc. These would appear an issue specific to
or only triggered on x86_64-apple-darwin10. I'll try the m64 multilib of the
i686-apple-darwin10 compiler next to see if it is specific to the native 64-bit
compiler on darwin10.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on x86_64-apple-darwin10 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #24 from howarth at nitro dot med dot uc dot edu  2009-09-06 00:29 -------
Okay, this issue also impacts the -m64 code generation of the
i686-apple-darwin10 target, I see the same set of failures as in Comment 12
with r151455. So this issue is darwin10 specific and isn't seen with either
x86_64-apple-darwin9 or i686-apple-darwin9 at -m64.


-- 

howarth at nitro dot med dot uc dot edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  GCC build triplet|x86_64-apple-darwin10       |*-apple-darwin10
   GCC host triplet|x86_64-apple-darwin10       |*-apple-darwin10
 GCC target triplet|x86_64-apple-darwin10       |*-apple-darwin10


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[dominiq at lps dot ens dot fr]

------- Comment #25 from dominiq at lps dot ens dot fr  2009-09-06 09:56 -------
> Dominique,
>     Also if you are bothering to run the test suite on i686-apple-darwin9
> periodically, you might as well shoot the results over to the gcc-testresults
> mailing list (since Apple has never set up a i686-apple-darwin9 regress
> server).

See

http://gcc.gnu.org/ml/gcc-testresults/2009-09/msg00494.html
http://gcc.gnu.org/ml/gcc-testresults/2009-09/msg00495.html
http://gcc.gnu.org/ml/gcc-testresults/2009-09/msg00502.html

I am curious to see how much it helps!-(


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #26 from howarth at nitro dot med dot uc dot edu  2009-09-06 15:07 -------
It duplicates my observations that this bug is only triggered on
*-apple-darwin10 at -m64.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[hjl dot tools at gmail dot com]

------- Comment #27 from hjl dot tools at gmail dot com  2009-09-06 15:37 -------
EH has been problematic on Darwin/x86, see PR 37012.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #28 from howarth at nitro dot med dot uc dot edu  2009-09-06 16:01 -------
Hopefully we can still use Apple's gdb to debug these EH issues on darwin10. If
I am reading...

http://gcc.gnu.org/ml/gcc-patches/2009-09/msg00074.html

correctly, the VTA merge may break the ability to debug optimized code but we
should be okay
debugging test cases that compile at -O0, right?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #29 from howarth at nitro dot med dot uc dot edu  2009-09-08 13:16 -------
This should be a P1 because it involves regressions from gcc 4.4.1 on a primary
target (i686-apple-darwin10 at -m64).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[jakub at gcc dot gnu dot org]

------- Comment #30 from jakub at gcc dot gnu dot org  2009-09-08 14:34 -------
You are reading it incorrectly, already before VTA merge those 2 patches were
needed to debug 4.5 generated optimized code.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[rguenth at gcc dot gnu dot org]

------- Comment #31 from rguenth at gcc dot gnu dot org  2009-09-18 09:12 -------
P1 to attract darwin maintainers.  We'll downgrade this if not fixed before
a release candidate (and honestly we then should drop *-darwin from the list
of primary targets).


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #32 from howarth at nitro dot med dot uc dot edu  2009-09-18 16:57 -------
This issue could be a linkage issue as described here...

http://lists.cs.uiuc.edu/pipermail/llvmdev/2009-September/025894.html

and it may be that the linkage for 10.6 needs to be adjusted
in config/darwin.h.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #33 from howarth at nitro dot med dot uc dot edu  2009-09-18 18:02 -------
The comments in...

http://lists.cs.uiuc.edu/pipermail/llvmdev/2009-September/025898.html

also are useful in understanding the dimensions of the problem with darwin
>=10.6.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #34 from howarth at nitro dot med dot uc dot edu  2009-09-18 19:24 -------
If I understand those messages from llvm-dev properly, on Snow Leopard symbols
from libgcc_s are ignored and the unwinder is gotten from libSystem now. So
doesn't this mean we are just seeing the incompatibilities (code forking)
between
the system unwinder in darwin10 and that used in FSF gcc trunk? It seems for
darwin10 that we would have to either build against the system libgcc with
--with-slib or move the unwinder into a libgcc-ext (as was proposed in PR3988
for providing the emutls symbols). Otherwise the unwinder in FSF and system
libgcc_s is ignored in favor of that in libSystem in darwin >= 10.6, if I am
reading this correctly.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #35 from howarth at nitro dot med dot uc dot edu  2009-09-18 22:45 -------
We have two messages on llvm-dev which discuss the exact cause of the breakage
on darwin10 and possible fixes...

http://lists.cs.uiuc.edu/pipermail/llvmdev/2009-September/025908.html
http://lists.cs.uiuc.edu/pipermail/llvmdev/2009-September/025909.html

It appears the options are 1) not to add epilog unwind information for Darwin
or 2) have the gcc driver implicitly add -no_compact_unwind to the link line.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #36 from howarth at nitro dot med dot uc dot edu  2009-09-19 02:18 -------
I can also confirm that adding -Wl,-no_compact_unwind to the linkage eliminates
the error when the g++.dg/torture/stackalign/eh-vararg-2.C test cases is
executed.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #37 from howarth at nitro dot med dot uc dot edu  2009-09-19 06:17 -------
Executing...

make -k check RUNTESTFLAGS="--target_board=unix/-Wl,-no_compact_unwind"

shows that all of the g++ regressions caused by 147995 are eliminated.
Unfortunately, this approach plays havoc with the gcc.dg/pch and g++.dg/pch
testsuite with hundreds of new failures of the form...

FAIL: ./array-1.H -g (test for excess errors)
FAIL: g++.dg/pch/array-1.C -g
FAIL: g++.dg/pch/array-1.C -g assembly comparison

so it definitely would be better to try to fix this issue by not adding epilog
unwind information for Darwin instead. It should also be noted that are
darwin10, libgcc is subsumed into libSystem and the FSF libgcc is never
actually used for unwinding regardless linkage order for -lSystem and -lgcc.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #38 from howarth at nitro dot med dot uc dot edu  2009-09-19 16:31 -------
The solution we want to implement is described below...

-------------------------------------------------------
I dug into this.  Based on the .s files in bugzilla, the latest gcc is  
now adding dwarf unwind info to describe the function epilog.  If you  
run dwarfdump --eh-frame on the .o files made with the new compiler,  
you'll see extra dwarf unwind instructions at the end like:

                 ...
                 DW_CFA_advance_loc4 (64)  #<-- advance to near end of  
function
                 DW_CFA_restore (rbp)
                 DW_CFA_def_cfa (rsp, 8)
                 DW_CFA_nop
                 DW_CFA_nop

The linker's conversion to compact unwind "runs" the dwarf unwind info  
for a function and then records the state at the end.  Adding unwind  
info for the epilog breaks this.  In the long term, I can add  
heuristics to the linker to detect that what looks like unwind info  
for the epilog and stop processing the dwarf instructions.

The short term fix for gcc is to *not* add epilog unwind information  
for Darwin.

Epilog unwind information is never needed for exception processing.   
Its only use is for debugging or sampling when you want to  
asynchronously make a stack back trace.

-Nick
------------------------------------------------------


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #39 from howarth at nitro dot med dot uc dot edu  2009-09-19 22:47 -------
The patch...

Index: gcc/config/darwin.h
===================================================================
--- gcc/config/darwin.h (revision 151890)
+++ gcc/config/darwin.h (working copy)
@@ -372,7 +372,9 @@

 /* Machine dependent libraries.  */

-#define LIB_SPEC "%{!static:-lSystem}"
+#define LIB_SPEC "%{!static:\
+       %:version-compare(>= 10.6 mmacosx-version-min= -no_compact_unwind)  \
+       -lSystem}"

 /* Support -mmacosx-version-min by supplying different (stub) libgcc_s.dylib
    libraries to link against, and by not linking against libgcc_s on

eliminates the regressions from r147995 as shown in the testsuite results...

http://gcc.gnu.org/ml/gcc-testresults/2009-09/msg01761.html.

Still it would be better to suppress the epilog unwind information on darwin 
instead of resorting to  -no_compact_unwind which leaves binaries built under
darwin8/9 
with gcc 4.5 non-portable to darwin10 and later.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #40 from howarth at nitro dot med dot uc dot edu  2009-09-20 18:27 -------
Is this the location of the additional epilogue notes being added in r147995?

@@ -8637,7 +8757,17 @@
                                          + frame.nregs * UNITS_PER_WORD
                                         + frame.nsseregs * 16
                                         + frame.padding0));
-             emit_insn (gen_rtx_SET (VOIDmode, stack_pointer_rtx, tmp));
+             tmp = emit_insn (gen_rtx_SET (VOIDmode, stack_pointer_rtx, tmp));
+
+             gcc_assert (ix86_cfa_state->reg == stack_pointer_rtx);
+             if (ix86_cfa_state->offset != UNITS_PER_WORD)
+               {
+                 ix86_cfa_state->offset = UNITS_PER_WORD;
+                 add_reg_note (tmp, REG_CFA_DEF_CFA,
+                               plus_constant (stack_pointer_rtx,
+                                              UNITS_PER_WORD));
+                 RTX_FRAME_RELATED_P (tmp) = 1;
+               }
            }
        }
       else if (!frame_pointer_needed)

If so, we can't just apply an "ifndef __APPLE__/#endif" wrapper appropriately
there to suppress the addition of the epilog note? The question remains if the
new unwinder
will tolerate missing them. Remember that on darwin >= 10, libgcc is subsumed
into
libSystem and the libgcc built by FSF gcc is never actually used. So we have to
be
careful to not break darwin9 and earlier when suppressing the epilog notes for
darwin10.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[joseph at codesourcery dot com]

------- Comment #41 from joseph at codesourcery dot com  2009-09-20 21:04 -------
Subject: Re:  [4.5 Regression] major regressions on
 *-apple-darwin10 at -m64 caused by r147995

On Sun, 20 Sep 2009, howarth at nitro dot med dot uc dot edu wrote:

> If so, we can't just apply an "ifndef __APPLE__/#endif" wrapper appropriately
> there to suppress the addition of the epilog note? The question remains if the

__APPLE__ relates to the host and this issue relates to the target.  You'd 
at least be looking at a new target hook.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[pogma at gcc dot gnu dot org]

------- Comment #42 from pogma at gcc dot gnu dot org  2009-09-22 02:28 -------
Subject: Bug 41260

Author: pogma
Date: Tue Sep 22 02:28:19 2009
New Revision: 151960

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=151960
Log:
PR middle-end/41260
* gcc/config.gcc: Use darwin9.h and darwin10.h on darwin10 and later.
* gcc/config/darwin10.h: Add file to pass -no_compact_unwind on
LIB_SPEC for darwin10 and later since it always uses the unwinder in
libSystem which is derived from the gcc 4.2.1 unwinder.

Added:
    trunk/gcc/config/darwin10.h
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config.gcc


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #43 from howarth at nitro dot med dot uc dot edu  2009-09-23 02:38 -------
Fixed as of r151961.Tested on x86_64-apple-darwin10.


-- 

howarth at nitro dot med dot uc dot edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260

[Bug middle-end/41260] [4.5 Regression] major regressions on *-apple-darwin10 at -m64 caused by r147995

[howarth at nitro dot med dot uc dot edu]

------- Comment #44 from howarth at nitro dot med dot uc dot edu  2009-09-28 13:29 -------
This issue of r147995 breaking the darwin unwinder due to the new unwinding
information for the function epilog doesn't occur on Leopard. Leopard's linker
doesn't default to compact unwinding unlike Snow Leopard's linker which does.
While passing -no_compact_unwind on darwin10 works around the issue, I have
opened  radar Problem ID: 7256880 "compact unwinding crashes on unwind
information for function epilog in FSF gcc" in hopes that the unwinder in
libSystem on Snow Leopard can be fixed to not crash on this unexpected
unwinding information. The radar report includes binaries, preprocessed source
and assembly files for all of the failing g++ testcases as well as script to
quickly run them against the shared libraries from FSF gcc.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41260