This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Fix bootstrap failure on sparc/HPPA


> > It fails in the same place with today's CVS mainline.
> I am just going to investigate it.  I had to leave yesterday.
> Thanks for letting me know!
> 
> Honza

Hi,
the scenario what happends is that jump threading threads jump of basic block
that sets register X, since it probles X is set to proper value previously in
this path via libcse.  The destination of edge uses register X making register
X magically live in original block where it used to be killed by the threaded
block.

Fixed by deleting the notes, bootstrapped/regtested on i386 and I am going
to install it as obivous.

Sat Mar 16 15:06:05 CET 2002  Jan Hubicka  <jh@suse.cz>
	* cfgcleanup.c (cleanup_cfg): Fix updating of liveness.

*** /usr/src/cfg/egcs/gcc/cfgcleanup.c	Wed Mar 13 21:51:02 2002
--- cfgcleanup.c	Sat Mar 16 14:50:20 2002
*************** cleanup_cfg (mode)
*** 1768,1774 ****
  	  /* Cleaning up CFG introduces more oppurtunities for dead code
  	     removal that in turn may introduce more oppurtunities for
  	     cleaning up the CFG.  */
! 	  if (!update_life_info_in_dirty_blocks (UPDATE_LIFE_GLOBAL,
  						 PROP_DEATH_NOTES
  						 | PROP_SCAN_DEAD_CODE
  						 | PROP_KILL_DEAD_CODE
--- 1767,1773 ----
  	  /* Cleaning up CFG introduces more oppurtunities for dead code
  	     removal that in turn may introduce more oppurtunities for
  	     cleaning up the CFG.  */
! 	  if (!update_life_info_in_dirty_blocks (UPDATE_LIFE_GLOBAL_RM_NOTES,
  						 PROP_DEATH_NOTES
  						 | PROP_SCAN_DEAD_CODE
  						 | PROP_KILL_DEAD_CODE


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]