Re: libexecdir/datadir used incorrectly

[Aharon Robbins]

> > From: Alexandre Duret-Lutz <address@hidden>
> > Cc: address@hidden,  address@hidden
> > Date: Wed, 09 Feb 2005 21:45:16 +0100
> > 
> > The changes we are discussing presently attempt to promote some
> > sane uses of subdirectories in these two directories.  I can't
> > see how making such suggestions will make it harder to predict
> > where data files and subprograms end up.
>
> Some of the changes suggested in this thread would change the place
> where Gawk installs its data files (in datadir or in some subdirectory
> thereof, perhaps even in a machine- and version-specific
> subdirectory).  I'm glad that Aharon rejected those changes, but I'm
> concerned some other package could accept them.

I think Eli is referring to the idea of

        $(datarootdir)/package/machine-triple-version/pkg-version/

which GCC uses.

IMHO, this is appropriate for things where architecture and/or OS make
a difference, like GCC, and I guess Emacs.  It's not necessary for packages
where the data files are architecture and OS independent, like gawk. For
such packages, a directory like

        $(datarootdir)/package/pkg-version/

is not unreasonable.  In the case of gawk, because the files change so little
from release to release, and are usually usable by previous versions, I think
it's reasonable to use gawk's current scheme of

        $(datarootdir)/package/

I think Eli's concern is that the long directory name not get codified in
the GCS and end up being used by programs that don't need such an over-specified
location.  I would agree.

Arnold