Path: utzoo!mnetor!uunet!husc6!mit-eddie!uw-beaver!tikal!sigma!uw-nsr!uw-warp!dennis
From: dennis@uw-warp.UUCP (Dennis Gentry)
Newsgroups: comp.emacs
Subject: Re: Trying to install emacs on VMS 4.5
Message-ID: <277@uw-warp.UUCP>
Date: 15 Dec 87 18:22:00 GMT
References: <8712111835.AA15816@mitre-bedford.ARPA>
Organization: The Warp, Seattle, WA
Lines: 101
In-reply-to: lauri@MITRE-BEDFORD.ARPA's message of 11 Dec 87 18:35:27 GMT


Subject: Problems with VMS GNU Emacs 18.49 on VMS; some solutions


>I am trying to install GNU Emacs on a VAX 8700 and VAX 11/785 cluster
>running VMS 4.5.  The problem is when I execute [.src]build.com.  I
>get the following error message after it loads a couple of lisp files
>(it dies right after it loads simple.el):
>
>	Invalid read syntax:  "] in a vector"

My guess is that you copied the .ELC files, too, but they got
corrupted.  So when TEMACS tries to load "simple,", it finds the .ELC
and tries to use that, but it's corupted.  To fix this, you can delete
all your .ELC files, but if you do that, I think maybe you'll need to
double (or so) PURESIZE, which is #define'd in vms.h (after being initially
#define'd in config.h).  Also, be sure to remember which .ELC files
you've deleted.

Once you've got your emacs dumped, you can byte-compile all the .EL
files that should be .ELC files, and then drop PURESIZE back to its
original value and recompile TEMACS and redump it.  

>The only problems I encountered until this point are:
>
>	1.  The aforementioned bug check when compiling LREAD.C.  I 
>	used the /NOOPT qualifier to resolve that.

Yeah, that's the only way around the bug in the DEC C compiler
(both version 2.1 and version 2.3-23).  I suggest /debug/noopt.

>	2.  SETVBUF is undefined in module SYSDEP when I execute
>	LINK.COM.  I don't see how this would have any effect on my
>	system, but I could be wrong.

You're right, it doesn't have any effect on your system except to stop
it from linking.  To fix this problem, just #ifdef it out.  Someone
didn't realize that VMS C's stdio defines _IOFBF, just like a "newer
USG system," but VMS C doesn't have a setvbuf.  Just use the setbuf
(found on the next line) instead.  If you use "#ifndef VMS ... else
... " to fix this problem, you'll end up duplicating a line or two,
but some C compilers don't know about "#if !defined(_IOFBF) &&
!defined(VMS)", and you won't want to break it for them.

>Other than that, everything proceeded as normal thus far.  Has anyone
>else encountered this problem?  Any suggestions?  Thanks.

Another problem you'll likely find is that it hangs at baud rates
lower than 2400.  We fixed the problem by #ifdeffing out the the
special case stuff for < 2400 baud in dispnew.c.  (About line 550-570).

You might not encounter this problem, but we did on one system and not
on another.  Emacs was treating all files as if they were
write-protected.  We finally figured out that you need world read on
rightslist.dat.  Changing the protection to wo:re fixed the problem.
We have a couple of otherwise identical systems, and one seemed to
come with rightslist.dat as (sy:rwed,ow:rwed,gr:,wo:) and the other
with (sy:rwed,ow:rwed,gr:rwed,wo:re).  It was weird.

If they (or anyone) wants specific chunks of code, I'd be happy to
send them out.

Now for my questions:

Is the Free Software Foundation incorporating any of these fixes into
the distribution?  Like I said, I'd be happy to send them diffs or
whatever.

Hasn't anyone made mail sending (and reading) work under VMS GNU Emacs
yet?  How about dired?  If this stuff hasn't already been done, I'll
start working on it shortly.  If you have any problems getting GNU
Emacs working properly under VMS, let me know.  I want to get it
working correctly (and more completely).  Of course, I'd rather be
running GNU Emacs under any Unix (or better yet under GNU OS!) but I,
like many others, am stuck with VMS.

Is anyone working on porting GNU CC to the AT&T 3b2?

Can anyone tell me the details of acquiring a runnable copy of GNU CC
for VMS?  I snarfed the regular (unix) distribution from prep, but
can't even compile the preprocessor because the DEC C compiler chokes
on a large-ish #define in config.h.  And the Makefile is complicated
enough that I'm not looking forward to doing all the compiling by
hand.  I read a fairly impressive blurb about the VMS version, but
I've never heard anything about it since then.  I can even afford to
pay(!) for an official distribution tape, but when I asked at the FSF,
they told me that an official tape was about a month away (two weeks
for whoever wrote it to get it to them, and two weeks for them to make
a tape).

If anyone could tell me who the third party was, or how to get a copy,
I would appreciate it.  Good luck on the port, and thanks for any
answers.

Dennis.
-------
arpa:   uw-nsr!uw-warp!dennis@beaver.cs.washington.edu
usenet: {ihnp4|decvax|...}uw-beaver!uw-nsr!uw-warp!dennis
-- 
Dennis.
-------
arpa:   uw-nsr!uw-warp!dennis@beaver.cs.washington.edu
usenet: {ihnp4|decvax|...}uw-beaver!uw-nsr!uw-warp!dennis