RTL_FLAG_CHECK patch

[law@redhat.com]

 In message <20020510.121658.26959745.davem@redhat.com>, "David S. Miller" 
writes:
 >    From: "John David Anglin" <dave@hiauly1.hia.nrc.ca>
 >    Date: Fri, 10 May 2002 15:26:00 -0400 (EDT)
 >    
 >    Hit this:
 >    
 >    ./xgcc -B./ -B/opt/gnu/hppa2.0w-hp-hpux11.11/bin/ -isystem /opt/gnu/hppa2
 > .0w-hp-hpux11.11/include -isystem /opt/gnu/hppa2.0w-hp-hpux11.11/sys-include
 >  -O2  -DIN_GCC    -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-pro
 > totypes -isystem ./include  -fPIC -g  -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED 
 >  -I. -I. -I../../gcc/gcc -I../../gcc/gcc/. -I../../gcc/gcc/config -I../../gc
 > c/gcc/../include  -DL_divdi3 -c ../../gcc/gcc/libgcc2.c -fexceptions -fnon-c
 > all-exceptions -o libgcc/./_divdi3.o
 >    ../../gcc/gcc/libgcc2.c: In function `__udivmoddi4':
 >    ../../gcc/gcc/libgcc2.c:64: RTL flag check: INSN_ANNULLED_BRANCH_P used w
 > ith une
 >    xpected rtx code `insn' in mark_set_resources, at resource.c:738
 > 
 > Can you print out the RTX it is looking at when this triggers?
 > 
 > There is a lot of code which assumes that the first instruction in a
 > delay slot sequence emitted by reorg is a branch.  Aparently PA
 > ends up putting something different there or the port is using
 > sequences for something outside of reorg.
It is possible to have a normal INSN with a delay slot on the PA.

jeff