"Sander Striker" <[EMAIL PROTECTED]> writes: > > From: [EMAIL PROTECTED] > > [mailto:[EMAIL PROTECTED]]On Behalf Of Jeff Trawick > > Sent: 17 April 2002 19:25 > > > "Sander Striker" <[EMAIL PROTECTED]> writes: > > > >>> 1) always do libtool --install > >> > >> +1. > > Can we agree on this? Since we are using libtool throughout our > entire build it seems inconsistent to bypass it here.
yes, we can agree on this... I'm planning to work on this tonight and in the morning. If you'd rather proceed yourself, that's fine, and I can test on AIX, Tru64, and HP-UX and assume that any oddities on other platforms are represented by what happens on those three platforms. > >>> 2) look in installed .la file for the name that must be passed to > >>> dlopen and rename that file in the target directory to what we > >>> think a DSO ought to be named > >> > >> But we only need to do that for AIX, Tru64 and HP-UX. > > > > I don't really believe that :) libtool has as many nuances as there > > are operating systems. It seems easy enough to look in the .la file > > and if the name to dlopen is different than what we want, rename what > > it got installed as. > > Can you give me some examples? I am totally in the dark here and > some input on what happens could possibly enable me to help solve > the problem. Here is what gets installed for me with libtool 1.4.2 on AIX: -rw-r--r-- 1 trawick staff 62568 Apr 18 12:53 ../../built/modules/libmod_foo.a lrwxrwxrwx 1 trawick staff 25 Apr 18 12:53 ../../built/modules/libmod_foo.so -> libmod_foo.so.0.0.0 lrwxrwxrwx 1 trawick staff 25 Apr 18 12:53 ../../built/modules/libmod_foo.so.0 -> libmod_foo.so.0.0.0 -rwxr-xr-x 1 trawick staff 71968 Apr 18 12:53 ../../built/modules/libmod_foo.so.0.0.0 -rw-r--r-- 1 trawick staff 792 Apr 18 12:53 ../../built/modules/mod_foo.la These variables in the .la file look interesting: # The name that we can dlopen(3). dlname='libmod_foo.so.0.0.0' # Names of this library. library_names='libmod_foo.so.0.0.0 libmod_foo.so.0 libmod_foo.so' > Also, is this solvable in libtool itself? IOW, can we come up with > a patch to libtool that solves the problem? sure, somebody could come up with a patch :) > >>> 3) since this design does libtool --install for every DSO, we need to > >>> work hard to zap the extra system-dependent libtool-created > >>> files/symlinks in the target directory > >> > >> But that's not too hard. We can zap everything not ending with .so, > >> right? > > > > sounds right... and you have to account for a possible "lib" prefix > > on some of the files > > How did the lib prefix get there in the first place? (does that happen > on AIX, Tru64 or HP-UX?). libtool decided that the lib prefix was appropriate (and I expect that it is correct, given the assumption that the library is for linking against) . > If you pass -module to libtool you can name > it anything you like; like we do for our modules. We're already using -module. Are we using it incorrectly? -- Jeff Trawick | [EMAIL PROTECTED] Born in Roswell... married an alien...
