Re: nt/makefile.w32-in

[Eli Zaretskii]

> Date: Fri, 22 Dec 2006 13:07:03 +0200
> From: Eli Zaretskii <address@hidden>
> Cc: address@hidden, address@hidden, address@hidden
> 
> > Date: Thu, 21 Dec 2006 23:48:44 +0100
> > From: Lennart Borgman <address@hidden>
> > CC:  address@hidden,  address@hidden,  address@hidden
> > >
> > > After the release, I will convert gmake.defs and nmake.defs to DOS EOL
> > > format.  I don't want to touch them now, as no one complained that it
> > > interferes with the build.
> > 
> > I guess you are pretty sure that this is the right thing to do. Why not 
> > do it now then?
> 
> Because this has been in the code base since December 2001, and no one
> complained.  I want to avoid introducing new problems for which we
> don't have enough testing time before the release.

Actually, the plot thickens.  It turns out Andrew Innes reverted back
to Unix EOLs a short while after the change to DOS EOLs:

    2000-09-17  Andrew Innes  <address@hidden>

            * gmake.defs: Revert to Unix line endings.

            * nmake.defs: Revert to Unix line endings.

    2000-09-14  Andrew Innes  <address@hidden>

            * makefile.w32-in: Revert to Unix line endings.

    2000-09-03  Andrew Innes  <address@hidden>

            * makefile.w32-in: Change to DOS line endings.


What happened next is two things: on 2001-09-25 the Emacs 21 release
branch was cut, and on 2001-12-03, nt/makefile.w32-in was checked in
on HEAD with this change:

            (bootstrap-nmake):
            (bootstrap-gmake): Extend bootstrap process to first do
            bootstrap-clean in lisp dir and rebuild the DOC file.
            (bootstrap): Do a "normal" make after the bootstrap work.

(which btw is not in ChangeLog).  This changed the file back to DOS
EOLs as a side effect, evidently by accident.

As a result, all Emacs 21.x releases were shipped with
nt/makefile.w32-in that had Unix EOLs, while in CVS we have DOS EOLs
since December 2001.

So, on the one hand, this makes a very strong argument for changing
the EOLs back to Unix, because all users of Emacs 21.x who built it
did that with Unix EOLs, but OTOH the experience of CVS usage tells us
that the current situation did not cause any trouble since 2001-12.

It is clear that the file should be reverted to Unix EOLs, but now I'm
unsure what, if anything, should we do before the release.

Thoughts?