Re: More problems with `make -n' in automake-generated rules.

[Ralf Wildenhues]

Hi Stefano,

* Stefano Lattarini wrote on Wed, Nov 03, 2010 at 06:30:34PM CET:
> On Monday 01 November 2010, Ralf Wildenhues wrote:
> > I noticed more issues with automake-generated rules and `make -n':
> > 
> > 1) The solutions documented in the `Multiple Outputs' node are not safe
> > for use with `make -n'.
> > 
> > 2) Consequently, the lisp rules are broken, but also the Yacc, Vala, and
> > config.h rules in some cases.
> > 
> > 3) The rules to update Makefile, but also those to update and
> > Makefile.in, are broken in some circumstances, too.

> > I'm not sure how useful it is to fix (3).  It is not easy as a user to
> > get GNU make to not update any of the dependencies of the Makefile file,
> > thanks to its remaking feature (info make "Remaking Makefiles").  I'll
> > reply with a patch for the 'Makefile' rule, but in order to expose that
> > bug, you need to use something like this in a subdirectory of a package:
> >   make -n Makefile AM_MAKEFLAGS="-n Makefile"
> > 
> > I don't think users go to this extent just to have `make -n' work, and
> > they definitely won't get the above right on the first try; but then the
> > rebuild will already have kicked in, making the issue moot for the
> > second try.
> FWIW, I agree that (3) is a minor problem.

I'm not going to push the patch for (3), I never intended to.

> > Before applying this (to maint, probably) I would appreciate if someone
> > could look over it to make sure the patch looks sane.  Thanks.
> I didn't spot any obvious error in the "meat" of the patch.  Just a couple
> of nits w.r.t. the test cases...

I agree with all of your nits for the first patch, and have pushed it
after fixing them.

Thanks for the review,
Ralf