PATCH resource.c/jump.c enable-checking failures

Jeffrey A Law law@cygnus.com
Fri Feb 25 21:21:00 GMT 2000


  In message <3884A653.33DA170@rcp.co.uk>you write:
  > Hi
  > 
  > patch #1 for resource.c which replaces the use of SET_DEST (...)
  > with XEXP(..., 0). SET_DEST was being used on a RTL node which was NOT
  > a SET or CLOBBER, this causes enable checking to fail.
  > 
  > ChangeLog
  > 
  > 	* resource.c (mark_referenced_resources): Changed use SET_DEST (...)
  > 	to XEXP (..., 0) on RTL nodes which are not SET or CLOBBER.
Thanks.  I installed this patch.

  > Patch #2 for jump.c which corrects enable checking failure in returnjump_p 
  > ().
  > returnjump_p is passed arbitary RTL nodes such as NOTES, SYMBOL_REFS, etc. 
  > Applying PATTERN () to these nodes is meaningless.
  > 
  > ChangeLog
  > 	* jump.c (returnjump_p): Add check for JUMP_INSN
I'm less sure about this one.  Under what conditions were we passed stuff
like NOTEs?


Every case in the sources I could find always verifies that it's passed
an INSN of some kind, never a NOTE, BARRIER or anything else that doesn't
have a pattern.

It is possible that originally the call to returnjump_p in emit-rtl.c passed
it random stuff (that code has been hacked on a lot recently).  But it now
verifies that it's passing a insn of some kind.


jeff



More information about the Gcc-patches mailing list