From: locke AT mcs DOT net (Peter Johnson)
Newsgroups: comp.os.msdos.djgpp
Subject: Re: NT4 DPMI BUG (303h, Allocate Real-Mode Callback)--ATTN: DJ Deloie, Shawn Hargreaves, Eli Zaretskii
Date: Fri, 3 Dec 1999 23:59:24 -0600
Organization: BiLogic Productions
Lines: 25
Message-ID: <MPG.12b269b4c39b61ed9896d5@news.cso.uiuc.edu>
References: <MPG DOT 12b2640e2b0657f09896d4 AT news DOT cso DOT uiuc DOT edu>
NNTP-Posting-Host: isr4186.urh.uiuc.edu
X-Newsreader: MicroPlanet Gravity v2.12
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

In article <MPG DOT 12b2640e2b0657f09896d4 AT news DOT cso DOT uiuc DOT edu>, locke AT mcs DOT net 
says...
> As the location of data variables in DJGPP is obviously not controllable 
> at build time, I devised the following workaround that should probably be 
> included in src/libc/dpmi/api/d0303.s:
> 
> 1) Get a selector from DPMI (function 0000h)
> 2) Get the base address of the DS selector (DPMI function 0006h)
> 3) Set the base address of the selector allocated in 1) to the value 
> found in 2) plus the offset of the registers variable (ARG2).
> 4) Set the limit of the selector allocated in 1) to 32h (the size of the 
> dpmi registers structure).
> 5) Pass the created selector and offset 0 as ES:EDI to 303h
> 
> Obviously, this selector should be saved somewhere so d0304.s can free 
> it.

Also, it would make sense to use this workaround only if the offset is 
>64k (otherwise its just a waste of selectors).

-- 
Peter Johnson             locke AT mcs DOT net
:Windows: Where do you want to go today?
:Linux:   Where do you want to go tomorrow?
:FreeBSD: Are you guys coming or what?