gcj static builds are failing

Wed Jan 10 16:17:00 GMT 2007

Tom Tromey writes:
 > >>>>> "Andrew" == Andrew Haley <aph@redhat.com> writes:
 > 
 > Andrew> At the moment we determine the availability of libmagic at libgcj
 > Andrew> build time. This is wrong, because even if libmagic is present when
 > Andrew> gcj is built, it isn't necessarily present when libgcj apps are built,
 > Andrew> and vice versa.
 > 
 > I don't agree with this. I think it is safe to assume that whatever
 > we discover at configure time is true for the lifetime of the libgcj
 > install. I can't think of a case anywhere in GCC where we do
 > something different than this.
I don't think that's a reason not to do it. There's a push to
separate static libs into separate -static packages: that's a good
idea, and it'll happen sooner or later. So, to make the libgcj config
reasonably robust we'd have to test separately for libfile.so and
libfile.a for each multilib.
The "file" package in Fedora is being split into file, file-devel and
file-libs. I suppose we could make libgcj development dependent on
file-devel, but that's pretty pointless. Also, dynamically linking
against libfile is much more robust if the database format used by
libfile changes. So, it's a better solution in every way.
Andrew.