Re: more build sherrif-ery (and a touch of auto*)

[James Henstridge <james jamesh id au>]

Bryan Clark wrote:

> On Wed, 2004-11-10 at 11:52 -0500, Luis Villa wrote:
>  
>
> > If the common decision is 'port to 1.7 now', we could always re-break
> > gnome-common and use tinderbox to track the broken modules.
> >    
>
> We've got Daniel Reed [1], extreme auto-foo samurai, who would be able
> to bring up most modules to the latest auto-tools versions pretty
> quickly.  He would like to port the modules to 1.9 (current) if people
> don't have any disagreements with that.  It seems reasonable to me to go
> for the latest current release if we're going to be putting time into
> this at all, but what the hell do I know about auto-tools. :-)
>
> So if there are not complaints about the version bump I'd suggest the
> maintainers duck and cover for a flurry of incoming patches.
>  
From my experience so far, migrating modules from Automake 1.4 to a 
later version is a lot more difficult than going from something like 1.7 
to 1.9 (which usually involves making sure that $DESTDIR installs 
work).  There are enough interesting features in newer Automakes that 
getting rid of 1.4 would be useful when working on other bits of the 
build infrastructure.

If people do want to push forward to Automake 1.9, it might be worth 
reverting this gnome-common patch:
   http://bugzilla.gnome.org/show_bug.cgi?id=128691
Currently gnome-common is picking the *oldest* acceptable available 
automake, which is likely to get us into the same situation we were in 
with Automake 1.4.

If a package works with multiple automake versions, it is best to pick 
the newest release, since the older versions contain known bugs (based 
on the contents of the NEWS file).

James.