Re: A better autogen.sh

[Eli Zaretskii]

> Date: Wed, 16 Mar 2011 10:52:51 -0700
> From: Paul Eggert <address@hidden>
> CC: address@hidden, address@hidden
> 
> > I could understand an argument that supporting Unix-like platforms is
> > easier.  (And even the "easier" argument is IMO minor, looking at all
> > the stuff in lib/ that is needed to support those Unix-like non-GNU
> > platforms.)
> 
> But that's exactly the point.  We cannot avoid the need to
> have porting support.  But we can separate porting concerns
> from mainline development concerns.  The stuff in lib/ is separate,
> and people who aren't worried about porting to non-GNU platforms
> don't need to worry about lib/.

Not true.  Anyone who makes changes to the build process or any files
that are related to the build process _must_ worry about lib/, because
without understanding how it works and how it affects the rest of the
code it is very easy to break something, especially if some obscure
use case is involved that will not pop up until much later.  Moreover,
even code reading requires awareness of the stuff in lib/.  For
example, without knowing that there are wrappers for system headers
there, it is impossible to understand how certain parts of Emacs code
work on any given platform.

IOW, the lib/ stuff is anything but transparent.  Much less
transparent, in fact, than the Windows-specific Makefile templates.

> The Microsoft ports should be more like that: they should be
> separated from the mainline code, and kept in their own msdos/ and
> nt/ subdirectories, and ordinary Emacs development should not have
> to worry about them.

"Separated" does not mean "segregated" or "ostracized".  People who
happen to use and build the Windows port are responsible for
non-trivial contributions to Emacs development that are certainly no
less than yours, see the logs.  They don't deserve this kind of
attitude, under which it is okay to make their life miserable and
every other day break the port they use, in the name of lopsided
ideological arguments.

I cannot begin to express how disappointed I am to get this kind of
attitude from one of the most veteran members of the GNU project.

> Of course this is just a goal and practical compromises must be
> made in some cases.  Currently, though, there are too many instances
> of material in the mainstream repository areas only because of Microsoft,
> these instances make it harder to do mainstream development,
> and we should therefore encourage attempts to move this material
> into msdos/ and nt/.

"Too many" is a wild exaggeration.  I doubt that you'd be able to
point out more than 1 or 2 that are not clearly marked either by their
names or parent directory.

Anyway, this is not about moving stuff, but about removing it (or not
updating it, which is effectively the same).