------- Comment #10 from hjl dot tools at gmail dot com 2010-09-03 17:09
-------
(In reply to comment #8)
> (In reply to comment #5)
> > Hmhm, IVOPTs expands crap into arguments.
> >
> > Index: gcc/tree-ssa-loop-ivopts.c
> > ===================================================================
> > --- gcc/tree-ssa-loop-ivopts.c (revision 163817)
> > +++ gcc/tree-ssa-loop-ivopts.c (working copy)
> > @@ -1023,6 +1023,11 @@ find_bivs (struct ivopts_data *data)
> >
> > type = TREE_TYPE (PHI_RESULT (phi));
> > base = fold_convert (type, base);
> > + STRIP_USELESS_TYPE_CONVERSION (base);
> > + if (CONVERT_EXPR_P (base)
> > + && (POINTER_TYPE_P (TREE_TYPE (base))
> > + != POINTER_TYPE_P (TREE_OPERAND (base, 0))))
> > + continue;
> > if (step)
> > {
> > if (POINTER_TYPE_P (type))
> >
> >
> > fixes the testcase, I am bootstrapping and testing this on x86_64 as a
> > short-term fix but have to think about it some more.
> >
> > Can you check if 32bit bootstrap succeeds with that?
> >
>
> I can bootstrap C with this one on Linux/x86.
>
With default languages, I got one regression:
Executing on host: /export/build/gnu/gcc-32bit/build-i686-linux/gcc/xgcc
-B/export/build/gnu/gcc-32bit/build-i686-linux/gcc/
/net/gnu-6/export/gnu/import/git/gcc/gcc/testsuite/gcc.target/i386/pr38240.c
-ansi -pedantic-errors -S -o pr38240.s (timeout = 300)
/net/gnu-6/export/gnu/import/git/gcc/gcc/testsuite/gcc.target/i386/pr38240.c:
In function 'g':^M
/net/gnu-6/export/gnu/import/git/gcc/gcc/testsuite/gcc.target/i386/pr38240.c:8:21:
internal compiler error: in convert_move, at expr.c:326^M
Please submit a full bug report,^M
with preprocessed source if appropriate.^M
See <http://gcc.gnu.org/bugs.html> for instructions.^M
compiler exited with status 1
with revision 163817 plus the fix.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45519
