Re: automake less verbose (iter 3.1)

[Ralf Wildenhues]

Hello,

Let's move this to automake-patches, too.

* William Pursell wrote on Mon, Dec 15, 2008 at 09:19:38PM CET:
> 
> Thanks for this Jan, it is really nice functionality.  I don't
> know if this is a portability issue, but I think it would be nice
> to change $< to $? in this section:

Yes, $< is a portability problem; while $? should be portable,
it is probably too verbose, as Jan already mentioned.

> > +           'am__1verbose_CCLD_1   = @echo "  CCLD    " $@ "<-" $<;',
> > +           'am__1verbose_CXX_1    = @echo "  CXX     " $@ "<-" $<;',
> > +           'am__1verbose_CXXLD_1  = @echo "  CXXLD   " $@ "<-" $<;',
> 
> In the case with lots of dependencies, that might lead to excess
> verbage, but that's what V=0 is for, right?

I guess.  Another problem is that, in the case of lots of dependencies,
this might actually lead to overlong command lines.  Esp. on command
line length-challenged systems like w32, IRIX, Tru64 this can be a real
problem.

> I noticed that your patch modifies m4/Makefile.in.   Is that
> correct behavior?  I'm still unclear on the merits of putting
> Makefile.in in the repository at all, but I would think that
> its content in the repository should be automatically
> generated rather than patched.

Well, in general I think most projects have moved away from committing
generated files.  `info Automake CVS' has some discussion about the
issue.  For the Automake package itself, the policy so far has been to
commit the generated files (along with the commits that create the
changes) for a special reason: it forces you to look at the changes that
the patch causes on them.  This has helped find issues early before.
More generally, the tree should always be able to bootstrap itself,
otherwise a patch should not be applied; the committed generated files
easily show that bootstrapping works.

You don't need to post the changes to generated files in patches; I will
usually regenerate them before pushing a patch.

Cheers,
Ralf