re: CVS commit: pkgsrc

[matthew green <mrg%eterna.com.au@localhost>]

> On Sun, 30 Oct 2011, OBATA Akio wrote:
> 
> > On Sun, 30 Oct 2011 17:08:42 +0900, Steven Drake <sbd%netbsd.org@localhost> 
> > wrote:
> > 
> > > On Sat, 29 Oct 2011, OBATA Akio wrote:
> > > 
> > > > For me, failed to build dbm.so in lang/python27, if gdbm package is
> > > > installed
> > > > from pkgsrc
> > > > on NetBSD-5.1.
> > > 
> > > Any chance you have something like LIBRARY_PATH=/usr/pkg/lib in your
> > > environment?
> > 
> > No.
> 
> Dose dbm.so get built if the gdbm package is not installed?

i noticed that python27 fails to build on -current too.

what i saw is that dbmmodule.c does not include <ndbm.h>, and
thus it misses all the declarations of it.  there's no code in
the python27 package that checks for ndbm.h existing, so this
code ends up not being included:

        #if defined(HAVE_NDBM_H)
        #include <ndbm.h>

it does i think end up doing this code:

        #elif defined(HAVE_BERKDB_H)
        #include <db.h>
        static char *which_dbm = "Berkeley DB";

but <db.h> does not define eg DBM or the functions expected.

i hacked around it by #define HAVE_NDBM_H before the block.
(i just noticed that i probably have this, too, which is wrong
for me:
        static char *which_dbm = "GNU gdbm";  /* EMX port of GDBM */
)


one other problem is that the python build doesn't actually
fail when the compile fails.  it just keeps going and reports
that a module didn't build, but returns success.


.mrg.