Tom Lane wrote:
> Lamar Owen <lamar.owen@wgcr.org> writes:
> > Tom, try out a PPC build on this one. I know of one problem that I have
> > to fix -- postgresql-perl fails dependencies for libpq.so (I backed out
> > the patch to Makefile.shlib).
> The backend seems to build OK, but the build fails in interfaces/perl5
> because libpq-fe.h isn't found. The compiler is getting passed
> -I/usr/include/postgresql, which might work if I'd already installed
> the RPM, but that's tough when I haven't built it yet :-(
Well, actually, the perl5 interface gets built a total of three times.
:-/
Two times are done by the main Makefile system, and the third is done
manually to pass the right options. The -I _should_, on the third run
(an 'install' run), get passed $RPM_BUILD_ROOT/usr/include/postgresql.
The main Make system builds it once, then the install phase builds it
again to get the proper libraries (which, in the buildroot environment
get clobbered). I have to perform the third build (which bypasses the
GNUmakefile/Makefile.pl mechanism and executes the (already setup for
install phase) Makefile directly, with the proper options. Which of
course forces another whole build.
> I dunno how to get the RPM build process to bypass perl support, so I
> can't get any further than that ...
rpm --define 'perl 0' -ba .....
or
rpm --define 'perl 0' --rebuild ..... for that matter.
That portion needs a good cross-platform test too..... :-) I haven't
given it a thorough test yet -- was planning on doing that Saturday
morning/afternoon, as I was going to go through an double-check
everything, possibly releasing a -3 at that point if I have made any
substantial changes. I don't forsee any at this early date, but I did
want to do some extensive build testing and upgrade testing. With the
way it has to be done, it takes quite some time to thoroughly test.
Can you email me a build log (need stdout AND stderr)?
--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11