Re: AM_MAINTAINER_MODE

[Diego Elio Pettenò]

On 07/02/2013 19:47, Stefano Lattarini wrote:
> So you want to allow users to disable maintainer-mode rules in every
> package?

Yes. Where users here is "distribution packagers".

> Better risk an extra rebuild than to miss a required one IMVHO.  Or
> understand why timestamps get mangled, and fix that problem instead of
> its symptoms (i.e., unnecessary rebuilds, in this case).

Yes and no. In some cases, the problem we get is that the rebuild only
happens in some circumstances, and thus the developer is missing it, but
it happens on the users' systems, and then they report a bug that we
can't reproduce...

Basically, I want to have a build failure rather than a build that might
be wrong.

> I failed to understand what you're saying here, sorry.  Care to rephrase,
> or give an example?

I don't have an example at hand, but let's say this:

 - you got a package that for whatever reason is completely messed up if
generated with automake-1.12, but works fine with 1.9;
 - when I'm rebuilding it as part of an ebuild (Gentoo's spec files
equivalent, give or take), I declare WANT_AUTOMAKE=1.9;
 - but I'm not rebuilding it in the ebuild;
 - until I get a patch that I don't check thoroughly and messes up the
timestamps;
 - I still do not rebuild autotools in a controlled fashion;
 - automake triggers the rebuild, and rebuilds with 1.12;
 - I'm screwed.

Variations can happen if for instance the configure relies on a variable
that is not declared with AC_ARG_VAR (way too common).

Yes, it's all solvable with more attention to details and similar, but
since we care for stuff to at least behave, --disable-maintainer-mode is
much nicer _to us_.

-- 
Diego Elio Pettenò — Flameeyes
address@hidden — http://blog.flameeyes.eu/