Re: [Bug gcc/12279] Re: GCC 3.4-20030913 VAX lossage.

[Matt Thomas <matt at 3am-software dot com>]

On Monday, September 15, 2003, at 09:21 PM, Ian Lance Taylor wrote:

> Matt Thomas <matt@3am-software.com> writes:
>
> > I have a set of diffs that add PIC code generation to the GCC 3.4
> > VAX target.  However, due to a bug (which I submitted this weekend)
> > in casesi processing, I can't fully test these changes.  However,
> > as far as I can tell, these changes don't introduce any extra
> > regressions.
> >
> > I could submit them to gcc-patches (because I really want these
> > changes in GCC 3.4) but that doesn't seem right.  But I can't test
> > them and the bug is beyond my keen to fix.  (The bug happens even
> > at -O0 so I can't work around it at all).
> >
> > Any suggestions on how I should proceed?
>
> This patch fixes the VAX ICE.  It looks to me like the cfgcleanup.c
> portion of this patch
>
> 2003-09-15  Ian Lance Taylor  <ian@wasabisystems.com>
>
> 	* cfgcleanup.c (label_is_jump_target_p): Correct use of table
> 	returned by tablejump_p.

It helps but now the ICE has just moved down a few lines.

Program received signal SIGSEGV, Segmentation fault.
0x08307008 in label_is_jump_target_p (label=0x488a6988, 
jump_insn=0x488a7340)
    at /u2/toolchain/gcc/gcc/cfgcleanup.c:669
669             if (XEXP (RTVEC_ELT (vec, i), 0) == label)
(gdb) list
664           rtvec vec = XVEC (tmp,
665                  GET_CODE (PATTERN (tmp)) == ADDR_DIFF_VEC);
666           int i, veclen = GET_NUM_ELEM (vec);
667
668           for (i = 0; i < veclen; ++i)
669             if (XEXP (RTVEC_ELT (vec, i), 0) == label)
670               return true;
671         }
672
673       return false;
(gdb) print i
$1 = 0

--
Matt Thomas                     email: matt@3am-software.com
3am Software Foundry              www: http://3am-software.com/bio/matt/
Cupertino, CA              disclaimer: I avow all knowledge of this 
message.