Re: 23-gary-eliminate-double-configure-at-bootstrap.patch

[Peter O'Gorman]

Peter O'Gorman wrote:

> Gary V. Vaughan wrote:
>
> > This is a work in progress really.  I am trying to get libtool, when 
> > bootstrap from fresh CVS, to not reconfigure itself at the make step.  
> > This patch is almost there, but my head is about to explode with the 
> > multi-level quoting required.
> >
> > This is, however, a significant improvement over HEAD and generates a 
> > correct and complete libtool script (including tags) in a single 
> > pass.  I'd like to commit this patch as a milestone, and pick away at 
> > refactoring it into the new macros so that libtool can be remade from 
> > a single AC_CONFIG_COMMANDS tag.
> >
> > Comments?
>
>
> Well, I would suggest a branch, but if it continues to bootstrap 
> correctly, I see no need. Haven't tried it to verify that though :).

Having now actually tried this patch, I'll have to reverse myself. make 
check no fails on darwin (and likely loads of other machines) because 
some of the tests configure scripts attemt to run the generated libtool 
script to get at settings. This no longer works because libtool is not 
generated yet, it is generated at the same time as the Makefiles are by 
config.status, which makes running the libtool script in configure in 
order to change the generated Makefiles a noop (unless you run configure 
twice, but that is hardly an option). This is a major behavioral change 
which, in my opinion, should not go into HEAD yet.

Please put it in a branch.

How is one now meant to get at the libtool settings since libtool is not 
generated until configure has completed?

Peter
--
Peter O'Gorman - http://www.pogma.com