[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: libtool headaches

Hi Filip,

On Sat, 28 Apr 2001, Filip Van Raemdonck wrote:

> (Please cc: me on replies, I'm trying to subscribe but the mailing list
> seems to be slow in sending confirm requests)

> Hi,

> I am working on a package of fam (available from
> http://oss.sgi.com/projects/fam/), and I have some trouble wrt libtool.

> They use a selfcontained version of libtool in the build process, to build a
> shared library. The resulting library gets named libfam.so.0.0.0
> This seemed rather odd to me, as the source distribution is already at 2.6.4
> (it's an opensource release of an old SGI Irix tool), and I was correct in
> being suspicious: in a rpm spec file distributed with the sources there were
> commented out lines in the `%files' section that referenced a libfam.1.0 file
> (but a comment also seemed to indicate failure of correct soname generation
> on rpm systems).

> I suspect this is probably just a problem with the way libtool gets called
> (either incorrect or incomplete), but I'm not too good on reading
> Makefile.in's and apparently (to make things even more complicated) the
> libtool isn't even really distributed but gets generated from a file
> `ltconfig' in the top source directory.

> I'd really like some help to build the library with a correct major and minor
> version number. It's not that I can't build it or it doesn't work right now,
> but I fear severe breakage if one of the version numbers changes.

It's not out of the question that a library which is released as version 2.6.4
would still have an so version of 0.0.0; the major version of the library
soname changes any time the library interface changes, but the major version
of the /package/ would change at the author's discretion.  Still, even though
it doesn't need to match the release version 2.6.4, 0.0.0 does seem a little
low.  If you search through the makefiles for 'soname', do you find anywhere
that this argument is being passed to libtool (or supposed to be)?

Steve Langasek
postmodern programmer

Reply to: