Make world problem - apparent 'sh' partial commit

This came in last night during the 'make installworld' phase of make world
first showing up under /usr/src/games.  I lost that output, and now it
happens very early in bootstrap tools.

I'd presume a change to either '/bin/sh' or *.bsd.mk caused it.

------------------------------------------------------------------------
cc -O -pipe -I/usr/obj/usr/src/i386/legacy/usr/include -I/usr/src/usr.sbin/kbdcontrol    -c lex.c
cc -O -pipe -I/usr/obj/usr/src/i386/legacy/usr/include -I/usr/src/usr.sbin/kbdcontrol    -L/usr/obj/usr/src/i386/legacy/usr/lib -static -o kbdcontrol kbdcontrol.o lex.o -ll
sh /usr/src/tools/install.sh -s -o root -g wheel -m 555   kbdcontrol /usr/obj/usr/src/i386/usr/sbin
===> gnu/usr.bin/gperf
/usr/obj/usr/src/i386/usr/src/gnu/usr.bin/gperf created for /usr/src/gnu/usr.bin/gperf
Syntax error: "for" unexpected
*** Error code 2

Stop in /usr/src/gnu/usr.bin/gperf.
*** Error code 1

Stop in /usr/src.
*** Error code 1

Stop in /usr/src.
*** Error code 1

Stop in /usr/src.
--------------------------------------------

If this was due to a change in sh, I have a catch-22 in that the make world 
doesn't get that far to fix it :-)  (yes, I know I can likely build it 
individually, but things like this and unbootable kernels keep happening; 
these require the ability to cross-build under -stable to be kept at least
for a while (I noted a comment that appeared to be from 'ru' to the
effect that it may not be possible much longer...))  (and if 'make' fails
in the /bin/sh subdir itself, we have a *real* catch-22...)

Followup:
OK - an individual rebuild of sh (1) worked, and (2) fixed the problem.
Lucky this time...

This probably resulted from a more general problem - how to assure that
cvsup has brought in all dependent changes (or at least find out when it
has not); there is no way to do this currently, I know that usually this 
is handled out-of-band with HEADS-UP messages but I didn't see one here...
I presume that this is a more general property of cvs (e.g. lack of a
partial-commit flag) and not just cvsup.
I thought that one of the advantages of cvs over some of its competitors
was that partial commits were less likely...

-- Pete