Anyone knows about a testcase which was fixed by Wed Dec 24 18:05:13 1997 SPARC change?

[Mike Stump]

> Date: Mon, 16 Oct 2000 21:49:00 +0200
> From: Jakub Jelinek <jakub@redhat.com>
> To: Mike Stump <mrs@windriver.com>
> Cc: gcc@gcc.gnu.org

> But even then it would be needed for -mflat only, as register windows take
> care of the non-flat pic pointer.

Imagine the case where a symbol needs to be accessed on the far side,
from the pic register, but after the EH mechanism has redone the stack
pointer, and the frame pointer and the entire frame.  You discover
that the pic register needs reloading, again, on the far side.

One can tell if this is needed anymore by ensuring that that doesn't
happen anymore.  It did at one time.  I suspect the dwarf2 stuff can't
do this, one can verify that, and the conditionalize the code further
to exclude that case, if one wanted.