Re: automake 1.5

[Neil Spring <nspring@saavie.org>]

> Apparently automake is broken in sid, in that most packages are written
> for automake 1.4, but automake 1.5 is in sid, and backwards compatibility
> isn't considered a feature or a goal for automake. (See #114130, #114481,
> etc)
> 
> Given this, and that some 300 or so packages build-depend on automake
> in unstable, we obviously need to do something.

I would want to have a better understanding of the problem.

As it stands, we seem to know:
 - several packages break with v1.5
 - many packages (294) build-depend on automake.

I think deciding how to proceed would be easier if we
knew^1:
 1) how many of the 294 actually break 
 2) how many are:
   a) fundamental automake 1.5 vs 1.4 architectural changes
   b) automake 1.5 new bugs 
   c) upstream package bugs
 3) how many of the failures can't be fixed elegantly
   with Kevin's help to remove the automake dependency,
   or upstream (automake) bugfixes.

I looked over the koffice failure (one of those mentioned
in #114130), and automake complains about what seems to be
a legitimate problem: multiple definitions of blah_LDFLAGS
in apparently auto-generated Makefile.am files.  It smells
like a class 2C failure to me, which is a bug regardless
of which version of automake is used (it's apparently more
picky about it's input).

Automake v1.5 fixes some 1.4 bugs (I contributed to a
wishlist report) so I'd like to see it make progress.

-neil
[1] I know it's hard to gather this information.