Re: [PATCH 1/6] Add --gnulib-version and --news options to announce-gen.

[Ralf Wildenhues]

Hello Gary,

* Gary V. Vaughan wrote on Wed, Sep 01, 2010 at 08:09:32AM CEST:
> On 1 Sep 2010, at 12:25, Ralf Wildenhues wrote:
> > That didn't help to make it more readable, or cause less buggy code
> > though.
> 
> Well that's because the shared code in getopts.m4sh is still getting
> exposure.  I think that M4SH_GETOPTS is infinitely more readable and
> maintainable, and especially worth the effort of using and debugging
> if I am to contribute a variation to Autoconf's m4sugar libraries.

I don't think it makes sense to let the users of Libtool, nor its
developers, be test drivers for proposed features to Autoconf.  The
testing Autoconf code is best done extensively inside the Autoconf
testsuite.  For the benefit of avoiding repeated regressions, I have
added Libtool testsuite exposure to a couple of the issues I found, in
v2.2.10-70-g21cffd1 and in v2.2.10-73-g3078821.  But fundamentally, I
think that is not the right approach.

> Why maintain several copies of the same code if we can do it just
> once?

Because the previous code (in ltmain) was not broken?  Because the move
introduced a few regressions?

> >> Surely you're not suggesting that we continue to hand code, maintain,
> >> synchronize the option parsing loop in each of our scripts?
> > 
> > Well, bootstrap didn't need one so far, did it?  How much maintenance
> > does an option parsing loop need, once it is written?  I didn't have the
> > feeling that that was a biggie on our list before that.
> 
> Certainly not a biggie.  But after using the M4SH_GETOPTS generated
> bootstrap script on my gnulib branch for less than a week, going back
> to the under-featured master branch version is already painful.

OK, so which M4SH_GETOPTS-related feature is missing from bootstrap?
Do you propose have a generated bootstrap?

Thanks,
Ralf