cross build for RELENG8 on RELENG10 now fails

[Don Lewis]

On 10 Sep, Mike Tancsa wrote:
> When I upgraded our cross build server to RELENG10 from 9, I lost the 
> ability to build RELENG_8 images.  Is there some magic that needs to be 
> done in order to make this work on RELENG_10 ?
> 
> it dies at
> 
> ===> gnu/usr.bin/cvs/contrib (depend)
> ===> gnu/usr.bin/cvs/cvsbug (depend)
> ===> gnu/usr.bin/cvs/doc (depend)
> ===> gnu/usr.bin/dc (depend)
> rm -f .depend
> mkdep -f .depend -a    -I/crossbuilds/src/8/gnu/usr.bin/dc/../bc 
> -I/crossbuilds/src/8/gnu/usr.bin/dc/../../../contrib/bc/h 
> -DHAVE_CONFIG_H 
> /crossbuilds/src/8/gnu/usr.bin/dc/../../../contrib/bc/dc/array.c 
> /crossbuilds/src/8/gnu/usr.bin/dc/../../../contrib/bc/dc/dc.c 
> /crossbuilds/src/8/gnu/usr.bin/dc/../../../contrib/bc/dc/eval.c 
> /crossbuilds/src/8/gnu/usr.bin/dc/../../../contrib/bc/dc/misc.c 
> /crossbuilds/src/8/gnu/usr.bin/dc/../../../contrib/bc/dc/numeric.c 
> /crossbuilds/src/8/gnu/usr.bin/dc/../../../contrib/bc/dc/stack.c 
> /crossbuilds/src/8/gnu/usr.bin/dc/../../../contrib/bc/dc/string.c 
> /crossbuilds/src/8/gnu/usr.bin/dc/../../../contrib/bc/lib/number.c
> echo dc: /crossbuilds/obj/8amd64/crossbuilds/src/8/tmp/usr/lib/libc.a 
> /crossbuilds/obj/8amd64/crossbuilds/src/8/tmp/usr/lib/libm.a >> .depend
> ===> gnu/usr.bin/dc/doc (depend)
> ===> gnu/usr.bin/dialog (depend)
> rm -f .depend
> mkdep -f .depend -a     /crossbuilds/src/8/gnu/usr.bin/dialog/dialog.c
> echo dialog: 
> /crossbuilds/obj/8amd64/crossbuilds/src/8/tmp/usr/lib/libc.a 
> /crossbuilds/obj/8amd64/crossbuilds/src/8/tmp/usr/lib/libdialog.a 
> /crossbuilds/obj/8amd64/crossbuilds/src/8/tmp/usr/lib/libncurses.a >> 
> .depend
> ===> gnu/usr.bin/dialog/TESTS (depend)
> ===> gnu/usr.bin/diff (depend)
> patch -s -b .orig -o context.c < 
> /crossbuilds/src/8/gnu/usr.bin/diff/context.c.diff 
> /crossbuilds/src/8/gnu/usr.bin/diff/../../../contrib/diff/src/context.c
>    I can't seem to find a patch in there anywhere.
> *** Error code 2

I ran into this same problem.  The BSD patch in 10 interprets the -b
option differently than the GNU patch in 8 and 9.  I worked around this
by tweaking the Makefiles for diff, diff3, and sdiff to get rid of the
-b .orig arguments to patch, since that should be the default anyway.

I was cross building i386 on an amd64 host and ran into another problem
with the cross version of lint (actually lint1) core dumping.  So far I
haven't been able to track down the cause of that.