[Bug rtl-optimization/24626] [4.1/4.2 Regression] internal compiler error: verify_flow_info failed

[danglin at gcc dot gnu dot org]

------- Comment #42 from danglin at gcc dot gnu dot org  2006-01-14 17:58 
-------
The patch in #41 doesn't seem to be a complete fix.   I see the following
assembler code for testcase #2 on hppa-unknown-linux-gnu:

        .LEVEL 1.1
        .text
        .align 4
.globl F1
        .type   F1, @function
F1:
        .PROC
        .CALLINFO FRAME=64,CALLS,SAVE_RP,ENTRY_GR=3
        .ENTRY
        stw %r2,-20(%r30)
        bl F3,%r2
        stwm %r3,64(%r30)
        ldi 0,%r22
        copy %r3,%r26
        .CALL   ARGW0=GR
        bl $$dyncall,%r31
        copy %r31,%r2
        copy %r28,%r3
.L2:
        bl T,%r2
        copy %r3,%r26
        ldw -84(%r30),%r2
        copy %r3,%r28
        bv %r0(%r2)
        ldwm -64(%r30),%r3
        .EXIT
        .PROCEND
        .size   F1, .-F1
        .ident  "GCC: (GNU) 4.2.0 20060113 (experimental)"

We save register %r3, but it's not initialized.  


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=24626