reg-stack.c - a botched patch?

[Robert Lipe]

I just patched from 0917 to 0922.     Patch 2.4 emitted no errors.
Brik doesn't detect a problem:

$ brik -C < ,brik
2226235129b BAD ./gcc/config/i386/sco5.h
2647840030b BAD ./gcc/config/i386/t-sco5
 762059038b BAD ./gcc/crtstuff.c
 223664707b BAD ./gcc/defaults.h
2958649215b BAD ./gcc/cp/parse.c
2638283476b BAD ./gcc/ginclude/math-3300.h

The first two are becuase I'm running Kean's OpenServer multilib patch
and have applied my patch to rename section names to be COFF-capable.
The third and fourth are also artifacts of the COFF unwinding stuff 
with the SCO assembler.   I haven't touched the last two.

Notice that reg-stack.c doesn't appear in the list.

When building gcc/reg-stack.c, I see 

 make bootstrap
make CC="cc " libdir=/usr/local/lib LANGUAGES="c "
cc  -c  -DIN_GCC    -g  -DHAVE_CONFIG_H     -I. -I../../egcs-970917/gcc -I../../
egcs-970917/gcc/config ../../egcs-970917/gcc/reg-stack.c
"../../egcs-970917/gcc/reg-stack.c", line 1357: error: undefined symbol: x


If I add a line so that x is a register rtx, it then blows up a little
later in the same function:

"../../egcs-970917/gcc/reg-stack.c", line 1368: error: undefined symbol: pat

Was there perhaps an incomplete patch applied to the master?

RJL