egcs-2.93.10 ICEs in emit_swap_insn on x86 at -Os -funroll-lloops

[N8TM@aol.com]

In a message dated 3/8/99 2:04:11 PM Pacific Standard Time,
law@hurl.cygnus.com writes:

<< I'm testing a fix for this buglet.
  >>
Thanks.  BTW, the same bug remains in 2.93.11, but otherwise, the latest
snapshots appear to contain a good g77.   I have a case (LFK #8) where
-march=pentiumpro results in fewer spilled pointers than without the
-march=pentiumpro if that's of any interest.