[PATCH] Use reversed_comparison_code in sched-deps.c

[Richard Earnshaw]

graham.stott@btinternet.com said:
> This patch breaks bootstraps with RTL checking enabled on i686-pc-linux-gn
> u

> Your changes to get_condition trigget

> crtstuff.c: In function `frame_dummy': crtstuff.c:323: internal compiler
> error: RTL check: access of elt 2 of `eq' with last elt 1 in
> get_condition, at sched-deps.c:156

> This is the offending code

>    cond = XEXP (SET_SRC (pc_set (insn)), 0);
>    if (XEXP (cond, 2) == pc_rtx)
>      return cond; 

Argh!  Somehow I managed to run the tests with checking disabled.

> All,
> 
> Just a follow on while I still awake
> 
> Graham Stott wrote:
> > Richard,
> 
> > 
> >   cond = XEXP (SET_SRC (pc_set (insn)), 0);
> >   if (XEXP (cond, 2) == pc_rtx)
>                      ^ I suspect this should be 0
> >     return cond;
> > 
> 
> I'm trying an overnight bootstrap with that change.

No, it's looking inside the if_then_else too soon (I over optimized the 
code :-( )

Here's the patch I'm testing now (with checking back on this time)...

Sorry about that.

R.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: fltcmp-4.patch
Type: text/x-patch
Size: 1081 bytes
Desc: fltcmp-4.patch
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20040830/99defe8e/attachment.bin>