Re: missing dependency in a very unusual case

[Alexandre Duret-Lutz]

>>> "Jim" == Jim Meyering <address@hidden> writes:

 Jim> I'm not sure it's worthwhile, since this is such an
 Jim> unusual scenario,

Thanks for the report and the patch.

My initial reaction to this was that the issue was more general:
we would like to relink any target that has a @substitution@ in
its _LDADD/_LIBADD variable, whenever the Makefile (*) changes.

But on second thought, wouldn't the same argument also apply to
any flags variables, like CFLAGS?  Maybe @CFLAGS@ was changed
when the Makefile was regenerated.  Should we recompile
everything anytime a Makefile is regenerated just in case
@CFLAGS@ might have changed?

The answer is not obvious to me.  There are cases where I'm glad
that only few files were recompiled after I changed a configure
option.  There are also cases where I'd like everything to be
recompiled.

At least right now I can choose one of these behaviors by typing
either `./configure' or `make clean; ./configure'.  If
everything was systematically recompiled I wouldn't have the
choice.  So I tend to favor the current behavior as far as
CFLAGS and siblings are concerned.  

What does it means for _LDADD/_LIBADD?  I don't know.  Should
the current behavior be kept for consistency with *FLAGS, or
should Makefile changes cause the relink needed in your
scenario?  I don't have any preference here.


(*) this file is not necessarily named Makefile!
-- 
Alexandre Duret-Lutz