[Bug target/55659] New: [4.8 Regression] [SH] Build failure with ICE in dwarf2out_var_location, at dwarf2out.c:20748

[Bug target/55659] New: [4.8 Regression] [SH] Build failure with ICE in dwarf2out_var_location, at dwarf2out.c:20748

[kkojima at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55659

             Bug #: 55659
           Summary: [4.8 Regression] [SH] Build failure with ICE in
                    dwarf2out_var_location, at dwarf2out.c:20748
    Classification: Unclassified
           Product: gcc
           Version: 4.8.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: kkojima@gcc.gnu.org
                CC: jakub@gcc.gnu.org
            Target: sh*-*-*


sh4-unknown-linux-gnu fails to build during compiling libgcc2.c with

trunk/libgcc/libgcc2.c: In function '__absvsi2':
trunk/libgcc/libgcc2.c:234:1: internal compiler error: in
dwarf2out_var_location, at dwarf2out.c:20748
 }
 ^
0x82e4b56 dwarf2out_var_location
    ../../ORIG/trunk/gcc/dwarf2out.c:20744
0x83439bb final_scan_insn(rtx_def*, _IO_FILE*, int, int, int*)
    ../../ORIG/trunk/gcc/final.c:2215

It looks to start to fail after

r194392 | jakub | 2012-12-11 19:41:44 +0900 (Tue, 11 Dec 2012) | 10 lines

        PR middle-end/43631
        PR bootstrap/55615
        * var-tracking.c (emit_note_insn_var_location): If insn is followed
        by BARRIER, put note after the BARRIER.
        (next_non_note_insn_var_location): Skip over BARRIERs.
        (emit_notes_in_bb): If call is followed by BARRIER, put note after
        the BARRIER.

A reduced testcase for -g -O2 is

void
foo (void)
{
  bar ();
}

.236r.barriers dump looks like:

...
(call_insn/j 6 5 7 (parallel [
            (set (reg:SI 0 r0)
                (call (mem:SI (reg/f:SI 0 r0 [160]) [0 bar S4 A32])
                    (const_int 0 [0])))
            (use (reg:PSI 151 ))
            (return)
        ]) foo.c:4 325 {sibcall_valuei}
     (expr_list:REG_DEAD (reg:PSI 151 )
        (expr_list:REG_UNUSED (reg:SI 0 r0)
            (nil)))
    (nil))
(barrier 7 6 14)
(code_label 14 7 15 3 "" [0 uses])
(insn 15 14 16 (unspec_volatile [
            (const_int 2 [0x2])
        ] UNSPECV_CMPXCHG_2) 408 {align_log}
     (nil))
(code_label 16 15 17 2 "" [0 uses])
(insn 17 16 18 (unspec_volatile [
            (symbol_ref:SI ("bar") [flags 0x41]  <function_decl 0xb750c600
bar>)
            (const_int 0 [0])
        ] 4) 404 {consttable_4}
     (nil))
(insn 18 17 19 (unspec_volatile [
            (const_int 0 [0])
        ] 11) 409 {consttable_end}
     (nil))
(barrier 19 18 13)
(note 13 19 11 (expr_list:REG_DEP_TRUE (concat:SI (pc)
        (symbol_ref:SI ("bar") [flags 0x41]  <function_decl 0xb750c600 bar>))
    (nil)) NOTE_INSN_CALL_ARG_LOCATION)
(note 11 13 0 NOTE_INSN_DELETED)

Before the latest var-tracking.c change, NOTE_INSN_CALL_ARG_LOCATION
note has been put just after barrier 7 instead of barrier 19.  It seems
that the insns for constant pools for this target make the middle-end
confused.

[Bug target/55659] [4.8 Regression] [SH] Build failure with ICE in dwarf2out_var_location, at dwarf2out.c:20748

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55659

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-12-12 07:28:32 UTC ---
Please provide preprocessed testcase and gcc options, so that this can be
looked at from a cross-compiler.
Very likely md reorg pass does something wrong.

[Bug target/55659] [4.8 Regression] [SH] Build failure with ICE in dwarf2out_var_location, at dwarf2out.c:20748

[kkojima at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55659

--- Comment #2 from Kazumoto Kojima <kkojima at gcc dot gnu.org> 2012-12-12 07:34:20 UTC ---
(In reply to comment #1)
> Please provide preprocessed testcase and gcc options, so that this can be
> looked at from a cross-compiler.
> Very likely md reorg pass does something wrong.

A reduced testcase for -g -O2 is

void
foo (void)
{
  bar ();
}

I've got a similar failure for the unified arm-elf build:

/home/kkojima/kaz/xarm-eabi-combined/combined/libgcc/libgcc2.c: In function
'__eprintf':
/home/kkojima/kaz/xarm-eabi-combined/combined/libgcc/libgcc2.c:2054:1: internal
compiler error: in dwarf2out_var_location, at dwarf2out.c:20748
 }
 ^
0x82a0dba dwarf2out_var_location
    /home/kkojima/kaz/xarm-eabi-combined/combined/gcc/dwarf2out.c:20744
0x82a0dba dwarf2out_var_location
    /home/kkojima/kaz/xarm-eabi-combined/combined/gcc/dwarf2out.c:20632
0x831eb1b final_scan_insn(rtx_def*, _IO_FILE*, int, int, int*)
    /home/kkojima/kaz/xarm-eabi-combined/combined/gcc/final.c:2215

A reduced testcase for arm-elf with -g -O2 might be more handy:

--
struct _on_exit_args { };
typedef struct __sFILE __FILE;
struct _reent {
  __FILE *_stdin, *_stdout, *_stderr;
};
extern struct _reent *_impure_ptr ;
void
__eprintf (const char *string, const char *expression, unsigned int line,
       const char *filename)
{
  fprintf ((_impure_ptr->_stderr), string, expression, line, filename);
  abort ();
}

[Bug target/55659] [4.8 Regression] [SH] Build failure with ICE in dwarf2out_var_location, at dwarf2out.c:20748

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55659

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.8.0

[Bug target/55659] [4.8 Regression] [SH] Build failure with ICE in dwarf2out_var_location, at dwarf2out.c:20748

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55659

--- Comment #3 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-12-12 09:56:29 UTC ---
Author: jakub
Date: Wed Dec 12 09:56:22 2012
New Revision: 194442

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=194442
Log:
    PR target/55659
    Revert
    2012-12-11  Jakub Jelinek  <jakub@redhat.com>

    PR middle-end/43631
    * var-tracking.c (emit_note_insn_var_location): If insn is followed
    by BARRIER, put note after the BARRIER.
    (next_non_note_insn_var_location): Skip over BARRIERs.
    (emit_notes_in_bb): If call is followed by BARRIER, put note after
    the BARRIER.

    2012-12-06  Jakub Jelinek  <jakub@redhat.com>

    PR middle-end/43631
    * var-tracking.c (emit_note_insn_var_location, emit_notes_in_bb):
    Clear BLOCK_FOR_INSN on notes emitted in between basic blocks,
    don't adjust BB_END when inserting note after BB_END of some bb.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/var-tracking.c

[Bug target/55659] [4.8 Regression] [SH] Build failure with ICE in dwarf2out_var_location, at dwarf2out.c:20748

[jakub at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55659

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

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

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> 2012-12-12 10:01:57 UTC ---
Should be fixed now.

[Bug target/55659] [4.8 Regression] [SH] Build failure with ICE in dwarf2out_var_location, at dwarf2out.c:20748

[kkojima at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55659

--- Comment #5 from Kazumoto Kojima <kkojima at gcc dot gnu.org> 2012-12-12 11:18:39 UTC ---
arm-elf and sh4-unknown-linux-gnu are built successfully with
revision 194442.  Thanks!