Re: -fpic on powerpc is still not quite right.

[Jeffrey A Law <law@cygnus.com>]

  In message <199907171216.WAA01564@geoffk.wattle.id.au>you write:
  > 
  > The problem is that you can't change the stack configuration after
  > reload has finished eliminating the frame pointer, etc., registers.
  > 
  > Unfortunately, reload can still generate new memory references after
  > this point.
  > 
  > So with -fpic, every function must have a stack frame capable of at
  > least saving r30 (the PIC offset table register), even if it probably
  > won't need it.
  > 
  > Alternatively, we could teach find_reload about PIC offset table
  > registers...
  > 
  > I attach a patch.  I've checked it on the 19990716 prerelease.  It
  > causes no testsuite regressions.
  > 
  > It may be too late to include this in 2.95, but it should go on the
  > release branch.
  > 
  > -- 
  > Geoffrey Keating <geoffk@cygnus.com> <geoffk@ozemail.com.au>
  > 
  > ===File ~/patches/egcs-13.diff==============================
  > Fri Jul 16 23:43:36 1999    <geoffk@cygnus.com>
  > 
  > 	* config/rs6000/rs6000.c (rs6000_stack_info): Don't allow reload
  > 	to change the stack frame configuration.
Thanks.  Installed on both the release branch and the mainline sources.
jeff