[Bug target/25176] New: FAIL: Array_3 -O3 execution - bytecode->native test

[Bug target/25176] New: FAIL: Array_3 -O3 execution - bytecode->native test

[amodra at bigpond dot net dot au]

libjava/testsuite/libjava.lang/Array_3.java has failed on powerpc64-linux for
quite a while.  See
http://gcc.gnu.org/ml/gcc-testresults/2005-03/msg01215.html.  The problem is
that in

  static int baz ()
  {
    int[] x = (int[])null;
    int nn = x.length;
    return 5;
  }
...
    try
      {
        nn = baz ();
      }
    catch (NullPointerException _)
...

we have

00000000100010b0 <._ZN7Array_33bazEv>:
    100010b0:   7c 08 02 a6     mflr    r0
    100010b4:   e8 62 80 50     ld      r3,-32688(r2)
    100010b8:   f8 01 00 10     std     r0,16(r1)
    100010bc:   f8 21 ff 91     stdu    r1,-112(r1)      <-- new stack frame
    100010c0:   4b ff fc b9     bl      10000d78 <._init+0x68>
    100010c4:   e8 41 00 28     ld      r2,40(r1)
    100010c8:   39 20 00 00     li      r9,0
    100010cc:   38 21 00 70     addi    r1,r1,112        <- restore old frame
    100010d0:   38 60 00 05     li      r3,5
    100010d4:   80 09 00 08     lwz     r0,8(r9)         <-- segv here
    100010d8:   e8 01 00 10     ld      r0,16(r1)
    100010dc:   7c 08 03 a6     mtlr    r0
    100010e0:   4e 80 00 20     blr

Note that the faulting instruction has moved past the stack frame restore. 
This confuses the dwarf2 unwinder, because no dwarf2 description is emitted for
the function epilogue.  We either need a dwarf2 epilogue description or must
prevent any instruction that might cause an exception from being scheduled
after the stack frame restore.


-- 
           Summary: FAIL: Array_3 -O3 execution - bytecode->native test
           Product: gcc
           Version: 4.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: amodra at bigpond dot net dot au
GCC target triplet: powerpc64-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25176

[Bug target/25176] FAIL: Array_3 -O3 execution - bytecode->native test

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2005-11-30 02:26 -------
Confirmed, also happens on ppc-darwin.  (Note this is most likely on other
targets too, there is a bug about x86_64 and this IIRC to get libuwind to pass
its tests).


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
 GCC target triplet|powerpc64-linux             |powerpc64-linux, powerpc-
                   |                            |darwin
           Keywords|                            |EH, wrong-code
   Last reconfirmed|0000-00-00 00:00:00         |2005-11-30 02:26:24
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25176

[Bug target/25176] FAIL: Array_3 -O3 execution - bytecode->native test

[amodra at bigpond dot net dot au]

------- Comment #2 from amodra at bigpond dot net dot au  2005-11-30 13:07 -------
I had a little play at implementing unwinder info for the epilogue.  It's easy
to arrange for a "DW_CFA_def_cfa_offset: 0" to be emitted on the stack restore.
 However, often the function exit isn't emitted last.  Many functions have
blocks of code emitted to higher addresses past the exit.  So you need to emit
further dwarf2 info specifying the stack offset in these blocks.

It probably wouldn't be hard to teach dwarf2out how to handle this, by tracking
NOTE_INSN_EPILOGUE_BEG and any following NOTE_INSN_BASIC_BLOCK, but I'm
wondering whether it is worth the increase in .eh_frame size..


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25176

[Bug target/25176] FAIL: Array_3 -O3 execution - bytecode->native test

[rth at gcc dot gnu dot org]

------- Comment #3 from rth at gcc dot gnu dot org  2005-11-30 23:07 -------
Fixing the unwinder to handle epilogues is harder than you think, I think.
My guess is that it'll require completely rewriting the unwind generator.
It'll need to be CFG aware, which means it'll have to happen while the CFG
is still available and do some sort of notes to remember it.  Or perhaps
preserve the CFG until the end of compilation.  Which would require some
handling for the reorg pass -- perhaps simply regenerating the CFG.

Not that I want to discourage anyone from tackling this, but the only short
term fix is some sort of barrier to prevent scheduling the epilogue past 
any potentially trapping insns.


-- 

rth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rth at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25176

[Bug target/25176] FAIL: Array_3 -O3 execution - bytecode->native test

[christian dot joensson at gmail dot com]

------- Comment #4 from christian dot joensson at gmail dot com  2005-12-01 10:06 -------
Not 100% sure, but I do believe this has happend on sparc/sparc64 linux for a
quite some time...


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25176

[Bug middle-end/25176] FAIL: Array_3 -O3 execution - bytecode->native test

[amodra at bigpond dot net dot au]

-- 

amodra at bigpond dot net dot au changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |amodra at bigpond dot net
                   |dot org                     |dot au
                URL|                            |http://gcc.gnu.org/ml/gcc-
                   |                            |patches/2005-
                   |                            |12/msg00094.html
             Status|NEW                         |ASSIGNED
          Component|target                      |middle-end
   Last reconfirmed|2005-11-30 02:26:24         |2005-12-02 00:29:16
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25176

[Bug middle-end/25176] FAIL: Array_3 -O3 execution - bytecode->native test

[amodra at gcc dot gnu dot org]

------- Comment #5 from amodra at gcc dot gnu dot org  2005-12-02 06:15 -------
Subject: Bug 25176

Author: amodra
Date: Fri Dec  2 06:15:54 2005
New Revision: 107871

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=107871
Log:
        PR middle-end/25176
        * function.c (expand_function_end): Emit blockage for unwinder
        after return label.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/function.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25176

[Bug middle-end/25176] FAIL: Array_3 -O3 execution - bytecode->native test

[amodra at gcc dot gnu dot org]

------- Comment #6 from amodra at gcc dot gnu dot org  2005-12-02 06:16 -------
Subject: Bug 25176

Author: amodra
Date: Fri Dec  2 06:16:21 2005
New Revision: 107872

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=107872
Log:
        PR middle-end/25176
        * function.c (expand_function_end): Emit blockage for unwinder
        after return label.


Modified:
    branches/gcc-4_1-branch/gcc/ChangeLog
    branches/gcc-4_1-branch/gcc/function.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25176

[Bug middle-end/25176] [4.0 only] FAIL: Array_3 -O3 execution - bytecode->native test

[amodra at bigpond dot net dot au]

-- 

amodra at bigpond dot net dot au changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|FAIL: Array_3 -O3 execution |[4.0 only] FAIL: Array_3 -O3
                   |- bytecode->native test     |execution - bytecode->native
                   |                            |test
   Target Milestone|---                         |4.1.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25176

[Bug middle-end/25176] [4.0 only] FAIL: Array_3 -O3 execution - bytecode->native test

[amodra at gcc dot gnu dot org]

------- Comment #7 from amodra at gcc dot gnu dot org  2006-01-26 11:18 -------
Subject: Bug 25176

Author: amodra
Date: Thu Jan 26 11:18:39 2006
New Revision: 110251

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=110251
Log:
        PR middle-end/25176
        * function.c (expand_function_end): Emit blockage for unwinder
        after return label.


Modified:
    branches/gcc-4_0-branch/gcc/ChangeLog
    branches/gcc-4_0-branch/gcc/function.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25176

[Bug middle-end/25176] [4.0 only] FAIL: Array_3 -O3 execution - bytecode->native test

[amodra at bigpond dot net dot au]

------- Comment #8 from amodra at bigpond dot net dot au  2006-01-26 11:19 -------
Fixed


-- 

amodra at bigpond dot net dot au changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25176