On Thu, Feb 15, 2007 at 01:12:57PM -0700, Brook Milligan
wrote:
> Joerg Sonnenberger writes:
> > I strongly object such a fix.
>
> However, this leads me to ask two questions. First,
could you please
> explain your rationale for those of us who do not
necessarily
> understand why you object so strongly. Perhaps there
are obvious
> platform-dependent issues that I do not know about that
would lead to
> this objection.
For once, it is likely to trigger issues on Darwin/OS X. It
is also very
bad practise as it tends to hide the real issues, e.g.
missing functions
which got removed because they aren't "used"
anymore.
In fact, I want to start making correct linkage a hard
requirement for
pkgsrc.
> Second, could you please explain how to make the
following work
> correctly in the absence of this patch: a program loads
via dlopen() a
> plugin that in turn resolves symbols in libgdal.so.
I think you are confusing cause and action here. There's one
and only
one situation where it is acceptable to have unresolved
symbols in a
shared object: if those are intended to be provided by the
main program.
Arguably that is extremely poor style. If it is really
absolutely
necessary to do that, it can often be worked around in a
platform
specific way.
Independent of that, I don't think this is the culprit here.
The
gdal-lib package correctly builds a shared library on
DragonFly, so it
is unlikely to apply here. Can you check the build output
for actual
errors?
Joerg
|