Re: AM_PATH_LISPDIR and dist_lisp_DATA when emacs is not installed breaks

[Simon Josefsson]

Alexandre Duret-Lutz <address@hidden> writes:

>>>> "Simon" == Simon Josefsson <address@hidden> writes:
>
>  Simon> In earlier versions, when emacs wasn't found, elisp files were not
>  Simon> installed at all.  Using latest CVS, they are written into /.
>
> Ooops, I overlooked this.  I'm feeling that it was a bad
> idea to promote 
>   lisp_DATA=a.el 
> over 
>   lisp_LISP=a.el 
>   ELCFILES=
>
> I think it's best to revert this change (i.e., document ELCFILES
> again, and disallow lisp_DATA) and come back to what we had in
> prior versions.
>
> Another possibility would be to specialize lisp_DATA so that it
> does not install anything if $(lispdir) is empty, but I'm not
> fond of that as other _DATA variables do not behave this way.
>
> Do you agree?

I rather liked the lisp_DATA construct.  In any case, I think doing
what it did (i.e., no byte compilation) should be simple.  For many
(if not most) packages, byte compiling is only harmful, so the
developer should not have to go through weird workarounds to get that
working.  I consider ELCFILES a weird workaround.

Another alternative, that I prefer, would be to have lisp_DATA install
into $prefix/share/emacs/site-lisp/ if $(lispdir) is empty.  Then it
would consistent with other _DATA variables (unconditional install)
and simple to use.