delorie.com/archives/browse.cgi   search  
Mail Archives: djgpp-workers/1997/11/20/12:05:30

From: Andrew Crabtree <andrewc AT typhoon DOT rose DOT hp DOT com>
Message-Id: <199711201702.AA165755340@typhoon.rose.hp.com>
Subject: Re: LFN on NT
To: eliz AT is DOT elta DOT co DOT il (Eli Zaretskii)
Date: Thu, 20 Nov 1997 9:02:20 PST
Cc: djgpp-workers AT delorie DOT com
In-Reply-To: <Pine.SUN.3.91.971120121158.11842H-100000@is>; from "Eli Zaretskii" at Nov 20, 97 12:12 (noon)
Reply-To: andrewc AT rosemail DOT rose DOT hp DOT com

 
> Does this mean that you have given up on the problem with Clearcase
> volume with 8+3 names?  (Read: can I nuke all the mail we exchanged on
> the subject and which I'm keeping for future discussions with you?)
> Or are you still trying to see the short names at least?
Lets see.  I am still trying to get the short file names to work.
However, even if I can get everything working correctly, I still have
about 4000 non-lfn files that I would need to rename to 8.3 files
before things started working.  This I am not looking forward to, and 
also don't know if I can.  Some of them come from third parties, so 
when we get updates it would cause huge problems.  We are still mainly
a unix shop here, just myself and a few others are trying NT.  
So, I am lobbying for Rational to provide 8.3 file names from their
software.
 
> Can't that DLL export the same LFN API that Windows 9X supports?
I tried looking into it and I am not sure, with a tendency to lean towards no.
All of the documentation I see shows how to hook accesses to certain IO 
ports (like to provide soundblaster emulation or something), or to hook 
hardware interrupts.  I think the problem is that the Virtual Dos Manager
already hooks int 21, and am not sure how a user supplied driver would 
be able to chain into it.  This would be ideal though, from a DJGPP code
standpoint.  Other problems with this approach though, if it works, is 
that it is slower than calling directly into a dll (so says the docs), and 
it requires users to modify their registry and install the dll 
before hand (in the other case it is loaded on the fly).  I could investigate
more time into it, but given my limited debugging capability with NT 
I am not sure it is worth it.  We don't have a lot of NT machines, and I would 
need to install debug builds on two of them, and then set up the serial connection
for deugging.  I will try though, to see if hooking the interrupt itself is even 
possible. 

> That would make libc immediately compatible, with no need to add any code
> bloat. 
Code bloat should be minimal.  I can keep the register calling convention the 
same or mostly the same (we lose register AX in the process), so it should 
just be a conditional and then either an int 21 or the special bop instruction.

Andy


--
_______       ___________________________________________________________
           /                       	                  Andrew Crabtree 
          /                          	      Workgroup Networks Division 
         ____       ___  /                                Hewlett-Packard 
        /     /    /    /   		                    Roseville, CA 
     __/   __/    _____/                                     916/785-1675
                 /                           andrewc AT rosemail DOT rose DOT hp DOT com
___________   __/   _____________________________________________________

- Raw text -


  webmaster     delorie software   privacy  
  Copyright © 2019   by DJ Delorie     Updated Jul 2019