Mail Archives: djgpp/2008/12/30/23:30:48
Hi
On Dec 23, 12:50 am, Rugxulo <rugx DOT DOT DOT AT gmail DOT com> wrote:
>
> On Dec 19, 5:30 am, Eli Zaretskii <e DOT DOT DOT AT gnu DOT org> wrote:
>
> > Recently, I have resumed maintaining the DJGPP Emacs, and in
> > particular the CVS code now builds and works for me. Bootstrap (which
> > is necessary if you build the CVS version, see below) also works.
>
> "make bootstrap" is a little more time consuming than 22.3 because
> byte-compiling has to be done from scratch. But you only have to do it
> once. As Eli says, grab a (large) cup of coffee. ;-)
And apparently, if you get newer .el files from CVS, this will update
the .elc files:
cd lisp ; make cvs-update EMACS=../bin/emacs.exe
> > Kudos to Rugxulo <rugx DOT DOT DOT AT gmail DOT com> who helped me lately to make it
> > work reasonably well on Vista (my own experience is mainly on XP SP2).
>
> It would be nice to have more Vista testers, esp. since MS is pushing
> Vista so hard these days (and Win2k3, Win7, etc. share the same
> kernel). My handicap now is a wonky wifi card on my Vista laptop as
> well as (temporarily) disconnected / unconfigured router, so I'm using
> the XP computer these days for Internet (since I don't grok routers).
Now I'm just using a really long cord running to the router.
> > If there are still users on this forum who run DJGPP tools on DOS or
> > Windows platforms, I would encourage you to try the CVS version now.
>
> Cygwin's CVS works fine, for example.
Er, maybe not perfect, it seems to crap out occasionally due to CVS
backup renaming or whatever. Apparently, retrying several times until
it finishes (gets to vms/ and says "files not found") is needed, at
least for me.
> > This is because Emacs 23 will be released soon, and I would like to
> > have it tested by as many DJGPP users as possible, to find and weed
> > out any remaining problems. The development version is quite stable,
> > and can be used for everyday work without fear of frequent crashes.
Okay, but today via e-mail you said "soon" isn't like next week or
anything, so apparently there's no huge rush, which is good to know.
> > You can find the instructions for getting the development version of
> > Emacs from CVS here:
>
> > http://savannah.gnu.org/cvs/?group=emacs
"Software Repository":
cvs -z3 -d:pserver:anonymous AT cvs DOT savannah DOT gnu DOT org:/sources/emacs co
<modulename>
(where <modulename> is optional)
BTW, don't run CVS from within your potential EMACS/ subdir but from
its parent instead.
In other words, run CVS from within C:\SOURCE, not C:\SOURCE\EMACS
(else you'll get another EMACS/ subdir created).
> > After downloading the sources, read the file msdos/INSTALL for build
> > instructions.
>
The basic commands seem to be "config.bat --with-system-malloc msdos",
"make bootstrap", "make distclean", "make install CFLAGS=-gcoff -Os",
etc. But for now you need Bash for the lisp subdir (i.e. a stable
release like 22.3 needs no byte compilation since it comes like that
by default). And for me (on Vista SP1), besides needing SYSTEM_MALLOC
to work correctly, building *with* SYSTEM_MALLOC works much less
reliably than without (go figure). If you don't build with
SYSTEM_MALLOC, Vista won't correctly byte compile ZIRANMA.el,
configure LEIM correctly, load big files > 10 MB, etc.
Juan, DJ, Andris, what does this mean?
Without the following (i.e. overriding the CFLAGS), it seems to not
finish compiling correctly for me (crashes bash and/or make and/or
NTVDM):
bash ; make install CFLAGS="-gcoff -Os"
Does that mean there's a bug in bash, make, or both?
Honestly, though, it's fairly slow to compile, so either use 3.4.4 (or
use -O or -Os) to save time. Then again, it's the byte compiling that
takes the most time initially.
> There are still a few problems that you haven't really mentioned
> although I'm not sure if these potential fixes are reliably working or
> not:
>
> 1). "update" / UPDATE.EXE (when called by CONFIG.BAT) is flagged by
> Vista's UAC three times, so it may be easier to just change that to
> say "djupdate" in the .BAT and manually do "cd /dev/env/DJDIR/bin ;
> stubify -g djupdate.exe ; stubedit djupdate.exe
> runfile=update" (untested on Vista but should work, right???).
Doesn't work, "djupdate" is still flagged. And accepting the UAC
prompts is more than an annoyance. It (AFAICT) causes some files to be
Admin status, which prevents later modifications. However,
"djupdat" (no e) causes no issues. I had to right-click EMACS/ in
Explorer and change all permissions (in all subdirs of EMACS/) to fix
this.
P.S. Eli is using Make 3.79.2, Bash 2.04, GCC 3.4.3, DJGPP 2.03p2
while I'm using latest /beta/ versions (3.81, 2.05, 4.32, 2.04 beta),
if that makes a difference (doubtful).
- Raw text -