=?iso-8859-1?Q?"Felix=20K=F6nig"?= <felix.koenig@web.de> writes:
> gcc -I../../../include -I../../../backend -I/usr/local/include -O2 -I/usr/local/inc
> lude -DBUILDING_DLL=1 -Wall -Wmissing-prototypes -Wmissing-declarations -I../.. -c
> -o istrat.o istrat.c
> istrat.c: In function `OperatorRelationFillScanKeyEntry':
> istrat.c:494: `F_OIDEQ' undeclared (first use in this function)
> istrat.c:494: (Each undeclared identifier is reported only once
> istrat.c:494: for each function it appears in.)
F_OIDEQ (and a lot of other F_xxx macros) should be declared in
src/backend/fmgr.h, which is normally generated during build by
the shell script src/backend/utils/Gen_fmgrtab.sh. I speculate
that fmgr.h is completely hosed (perhaps empty) due to script
execution problems, and that this just happens to be the first
place in the compile that references an F_xxx macro.
Gen_fmgrtab.sh depends on a shell, awk, sed, and a bunch of other
stuff, so I wouldn't be too surprised if it fails under Win2000.
Not sure why Sasha is seeing a problem, though, unless he's trying
to build under Windows ...
regards, tom lane
