Mailing-List: contact cygwin-apps-help AT cygwin DOT com; run by ezmlm
Sender: cygwin-apps-owner AT cygwin DOT com
Message-ID: <3C146400.10504@ece.gatech.edu>
Date: Mon, 10 Dec 2001 02:28:00 -0500
From: Charles Wilson <cwilson AT ece DOT gatech DOT edu>
User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:0.9.4) Gecko/20010914
MIME-Version: 1.0
To: cygwin-apps AT cygwin DOT com
Subject: libtool packages 
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit

I've put three new packages here:
   http://www.neuro.gatech.edu/users/cwilson/cygutils/testing/

I am not yet ready to even ask for official inclusion.  This is just a 
kind of "here, heads up, what do you think?" message.

There's
   libtool-20010531-rc6.tar.bz2
   libtool-20010531-rc6-src.tar.bz2
   libtool-devel-20010531-rc6.tar.bz2
   libtool-devel-20010531-rc6-src.tar.bz2
   libtool-stable-1.4.2-1.tar.bz2
   libtool-stable-1.4.2-1-src.tar.bz2

Just like the auto* packages, the libtool-devel package installs into 
/usr/autotool/devel/, and the libtool-stable package installs into 
/usr/autotool/stable/.  The libtool package is a set of wrapper scripts 
that install into /usr/, and set the PATH and other ENV vars 
appropriately before exec'ing the "correct" real libtool.

libtool-stable:
   contains libtool-1.4.2 (mostly OOB, patches are documentation only)
   works (more or less).  Follow the goat book examples to build DLLs 
using this version.

libtool-devel:
   Robert Collin's hacked version.  (BTW, I recall that "edward 
tailbert" submitted a number of libtool fixes to cygwin@ and libtool@ 
lists back in March.  Some (I think) were actually folded into libtool 
cvs and are part of libtool-1.4official.  Do any of Robert's changes 
harken back to edward?)  Anyway, it is based off of libtool-cvs on 
20010531, with a 40k patch and then a bootstrap (redo the autoconf, 
automake, autoheader, etc).  Now that I've parsed out the minimum 
changes, I'll submit that to Gary Vaughan -- hopefully he'll find that 
easier to deal with than the earlier 400k patch....
   Anyway, this version seems to be able to build DLLs OOB, using the 
autoimport stuff.  (Of course, you have to re-libtoolize your target 
package with the new version to get that to work -- which may also 
entail updating configure.in to work with the new autoconf, and 
re-autoconf'ing and re-automaking...)

libtool:
   contains "libtool-scripts-20010531' (I'm basing the versioning of my 
-scripts- packages off of the cygwin -devel- version that they are 
supposed to work with.)  Again, the choice of whether to use -stable- or 
-devel- is based on configure.in: AC_PREREQ(X) X <= 2.13, stable; X > 
2.13 or nonexistant, devel.

Note that BOTH libtool-1.4.2 (1.4.1, 1.4) and hacked libtool do NOT use 
ltconfig.sh at all.  They only need ltmain.sh.  So, when 
re-libtoolizing, be sure to rm ltconfig.sh...

packaging:
   As with the auto* packages, the -devel- version installs its info 
files into /usr/info as well as /usr/autotool/devel/info/.  However, 
since libtool also ships a library, ltdl, I'm in a bit of a bind.

The static lib can *probably* stay where it is (usr/autotool/*/lib) and 
ltdlized packages will link to it okay.  (If not, most ltdlized packages 
ship with the source code anyway, and build it themselves locally if it 
"wants" to link statically.  Kinda like gettext/libintl).  However, both 
stable and devel versions ALSO ship a dll.  And, since the source code 
hadn't changed, the libtool version is 3:0:0 for both, so both the 
stable and devel packages contain a 'cygltdl-3.dll'.

But they are not interchangeable.  One was linked "the old way" with 
__declspec's and everything, while the other is a new-style 
autoimport/autoexport library.  (I *think* the new one can sub for the 
old one, but not vice versa)

Anyway, since they can't BOTH be in /usr/bin, and you DON'T want to set 
PATH to /usr/autotool/*/bin (kinda defeats the whole wrapper script 
idea) -- I arbitrarily put the -devel- version of the ltdl libs into /usr

Questions, comments, test reports?  They seem to work okay for my 
minimal tests...

--Chuck