Re: support standalone libltdl [libtool--gary--1.0--patch-23]

[Gary V. Vaughan]

Hallo Ralf!

Ralf Wildenhues wrote:
> * Gary V. Vaughan wrote on Tue, Apr 26, 2005 at 03:13:17PM CEST:
> > That done, a backport to branch-2-0 will require reverting to running
> > libltdl/configure as a subconfigure, otherwise we can't release it
> > bootstrapped with autoconf-2.59 and automake-1.9.5.
> >
> > HOWEVER: autoconf-2.59 still doesn't recognise Darwin's use of -lcrt2.o
> > in g77, so the fortran tests will always fail on Darwin.  Maybe we should
> > bootstrap even branch-2-0 with CVS autoconf to avoid that problem?
> 
> I think that bootstrapping with this bug fixed is a good
> idea.  It will not make any difference to all the packages that use
> autoconf themselves, however, so it won't gain as much as we hope.

Indeed, but I want to avoid repeated bug reports against the libtool testsuite
because of that bug in the autoconf we bootstrap with.

> My fear is that silent dependence on some CVS feature creeps in.

Agreed.

> That would mean big trouble, much bigger than the -lcrt2 bug.  If
> that were the case, I'd oppose to using CVS versions.

Good point.  Then we should provide patches against autoconf-2.59 and
automake-1.9.5 with the libtool-2.0 release for just the bugs we need fixed.
Then, if anyone wishes to reconfigure the tree they have the choice of
patching the released autotools, or installing CVS autotools before running
the libtool bootstrap script.

So far, I'd like to see the crt2 bug fixed, and my pending patches for this
changeset applied so that the differences between 2.0 and HEAD are minimised.

>>If we decide to do that, then it might ease our maintenance burden to
>>backport this patch more or less as is (without a subconfigure
>>hack)... the whole changeset is, after all, inspired by a bug in
>>branch-2-0.  Thoughts?
>  
> For this particular fix, it seems a good idea.

It seems you agree ;-)

> OTOH, if we decide to do this, it also means we have to support 1.5
> at least [until] Autoconf-2.60 and Automake-1.10 are out.

Why?  Libtool-2.0 will be perfectly useable in projects using ancient
autotools, it's just the libtool bootstrap that needs patched autotools.
We've taken great pains to make upgrading from 1.5 -> 2.0 as easy as possible,
so the response to bug reports against 1.5 should be: please upgrade to the
latest release, autoupdate your project, and try again.

>>After the backport (in whatever form it eventually takes), is there
>>anything else holding back a 2.0 release that I've forgotten?
>  
> For one: people need to be able to do compilation tests with libtool
> during `configure'.  Several people have complained about this.
> This is _the_ bug where I really don't know how to attack it.

Oh bugger.  I thought this was the last show stopper.  I *did* have some ideas
about how to solve that problem though.  I hope they will come back to me if I
go back through the list archives...

>>If not, I propose an alpha release, and a round of testing, with 2.0 a
>>few weeks later! :-D
> 
> I'll post a TODO update for HEAD soon, you may decide if there is
> anything else worth holding it up.  (Of course, I'll try to fix as
> many bugs from that list as possible in the remaining few weeks :-)

That would be a few weeks beyond getting a fix for the above into the tree, so
you have a while remaining ;-)

> I'd really like to at least get the forward ports of my recent
> branch-1-5 patches into that alpha, should it come.  OK?  :-)
> Otherwise, we will have to deal with all the Solaris trouble
> once more (the Sun Studio Fortran failures need both my boilerplate
> patches and a patch only in CVS Autoconf).

Sure.  Another patch against 2.59 to apply before bootstrapping the release
and to ship with the release incase people want to rebootstrap.  Perhaps we
should look at patches against released autotools in the debian and redhat
releases to see if there are more we should include?

>>Okay to commit (after autoconf & automake changes are in CVS)?
>  
> I for one need some time to digest this patch.  Any of my above
> thoughts are based on your comments only.

No problem.  That is why I posted early, rather than waiting until the
autoconf/automake patches are in.

I've noticed that I forgot to document the changes to LTDL_INIT, and update
NEWS.  I'll post those separately (for application to HEAD as one big patch).

Cheers,
        Gary.
-- 
Gary V. Vaughan      ())_.  address@hidden,gnu.org}
Research Scientist   ( '/   http://tkd.kicks-ass.net
GNU Hacker           / )=   http://www.gnu.org/software/libtool
Technical Author   `(_~)_   http://sources.redhat.com/autobook

signature.asc
Description: OpenPGP digital signature