Re: bug#9773: "make install-info" with DESTDIR writes unwanted usr/share/info/dir.gz

[Stefano Lattarini]

On Sunday 23 October 2011, Ralf Corsépius wrote:
> On 10/23/2011 01:26 PM, Stefano Lattarini wrote:
> > On Sunday 23 October 2011, Jonathan Nieder wrote:
> >> Stefano Lattarini wrote:
> >>> On Monday 17 October 2011, Jonathan Nieder wrote:
> >>
> >>>> +++ b/ChangeLog
> >>>> @@ -1,3 +1,16 @@
> >>>> +2011-10-17  Jonathan Nieder<address@hidden>
> >>>> +
> >>>> +        For Debian Bug#543992:
> >>>> +        * lib/am/texinfos.am (install-info-am): When deciding whether to
> >>>> +        add to an info/dir file, instead of checking for Debian
> >>>> +        install-info, check if DESTDIR is nonempty.
> >> [...]
> >>> I must say that I don't like this very much.  A better fix IMHO would
> >>> be to have an environment variable to inhibit the creation/updating of
> >>> the `${infodir}/dir' file
> >>
> >> Yes, that sounds sensible.  It would be better than checking for
> >> Debian install-info (because that was insane) and better than the
> >> current state (because it gives packagers on all distros an easy way
> >> to request inhibiting creation of the ${infodir}/dir file).
> >>
> >> Keying behavior on $(DESTDIR) left an unpleasant taste in my mouth,
> >> too.  Thanks for your thoughtfulness.
> >>
> > Glad we agree :-)
> 
> OK, your are forcing me to pronounce my disagreement.
> 
> Making package building sensitive to environment variables contradicts 
> the working principles of autoconf: Building packages is supposed to 
> examinine the system and to decide upon what to do.
>
So we should ignore CFLAGS and DESTDIR as well?  Speaking more seriously:
we *are* examinining the system to decide what to do ("run install-info,
if its available"), but also allowing the user to override this decision.
Which is absolutely in the autotools spirit ("the user is always right"),
and gives distro packagers an easy way to obtain what they need *today*.

> > Here is my proposed patch (for maint).  I will push in 72 hours if there
> > is no objection or review by then.
> 
> Then: Veto.
>
I'm sorry, but the only people I know with veto power on Automake commits
are its maintainer and its co-maintainers.  I was asking for reasoned,
justified objections; you failed to provide any so far.  The running
clock stands.

Stefano