Thu Apr 10 21:21:00 GMT 2003
> On Tue, Apr 08, 2003 at 08:15:37PM +0200, Jan Hubicka wrote:
> > I am getting quite nervous about keeping the
> > inconsistent asm statement in the chain. What about zapping all that
> > does not validate just after instantiate_virtual_regs?
> I thought we already did that... Certainly I think that's
> for the best. Does that affect whether the addressof changes
> are needed?
The attached patch avoids the crash as well. I am unsure on whether it
won't match on ASM statements that previously got magically trought the
compiler but I can not think about sane use of such trick.
OK for mainline/3.3?
Thu Apr 10 23:17:30 CEST 2003 Jan Hubicka <email@example.com>
* function.c (instantate_virtual_regs): Verify that all ASM statements
match after the virutal regs instantiation.
RCS file: /cvs/gcc/gcc/gcc/function.c,v
retrieving revision 1.415
diff -c -3 -p -r1.415 function.c
*** function.c 9 Apr 2003 06:32:21 -0000 1.415
--- function.c 10 Apr 2003 21:16:55 -0000
*************** instantiate_virtual_regs (fndecl, insns)
*** 3592,3597 ****
--- 3592,3603 ----
if (GET_CODE (insn) == CALL_INSN)
instantiate_virtual_regs_1 (&CALL_INSN_FUNCTION_USAGE (insn),
+ /* Past this point all ASM statements should match. Verify that
+ to avoid failures later in the compilation process. */
+ if (asm_noperands (PATTERN (insn)) >= 0
+ && insn_invalid_p (insn))
+ instantiate_virtual_regs_lossage (insn);
/* Instantiate the stack slots for the parm registers, for later use in
More information about the Gcc-patches