Arguments against including libtool with gnucash source?

Bill Gribble grib@linuxdevel.com
04 Nov 2001 18:33:38 -0600


On Sun, 2001-11-04 at 18:05, Dave Peticolas wrote:
> But will the ltmain.sh create libraries that can be opened by
> libtool 1.3.4? And will this mean libtool 1.3.4 users will now
> be subjected to the extremely large link times of libtool 1.4?

Obviously we need to make sure that things still work when the
"installed" version of libtool is 1.3.4.  I believe that there's no
problem; libtool docs explicitly discuss the situation where the
installed version of libtool is different from the one that you put into
the source. Remember, 'libtool' is just a script that passes the right
arguments to the linker; it doesn't really open libraries at all, and
once they have been created and installed, the .so libs are just plain
libs.  

I wasn't trying to take any position at all about what version of
libltdl to use, or whether we should include its source in CVS also
(that's another option to libtoolize).  That may be something that we
have to deal with also.  I just want to clean up the Makefiles and get
rid of the dependency on stuff being in ./.libs.  

As to the long link times.  This is very irritating, no doubt.  I think
the right answer is to see what we can do to fix that problem before we
force everybody onto libtool-1.4. I think we have a handle on how to fix
the problem ... which is one of the reasons I want to ship ltmain.sh
with the CVS source :) 

it seems that the "dependency_libs" that get built by 1.4 are hellishly
redundant, and that's what takes all the time.  We may be able to patch
ltmain.sh to run a simple uniqifying algorithm.  rlb has already found
just the thing in the gnome-config script.  It shouldn't be hard to
adapt. 

b.g.