Frame pointer elimination on ia32

[law@redhat.com]

In message <20020620191259.GM17171@atrey.karlin.mff.cuni.cz>, Jan Hubicka write
s:
 > > 
 > > I know this won't be popular, but omitting the frame pointer in leaf
 > > functions on ia32 simply isn't acceptable at the moment.  It breaks
 > > GDB pretty badly.  Tsk tsk Richard for not looking into the regresions
 > > that have been constantly reported by the tester for the last 3 weeks.
 > > 
 > > Once GDB is updated to deal with a frame-pointer-less leaf function, then
 > > we can safely turn this option back on.
 > 
 > With DWARF2 this should be done easilly now.  Perhaps we should pay
 > attention to that instead and default -fomit-frame-pointer as I've
 > memtioned in the other post.  I hope that major issues in DWARF2
 > unwinding are slowed now.
Agreed, but until someone actually writes the GDB side and gets it integrated
we should not be eliminating the frame pointer on i686-pc-linux.gnu or
other targets were debugging frame-pointerless functions doesn't work.

jeff