Date: Mon, 14 Jul 1997 14:19:18 +0000
From: Bill Currie <billc AT blackmagic DOT tait DOT co DOT nz>
Subject: Re: Q: interrupts and segment changes.
To: Shawn Hargreaves <Shawn AT talula DOT demon DOT co DOT uk>
Cc: djgpp AT delorie DOT com
Reply-to: billc AT blackmagic DOT tait DOT co DOT nz
Message-id: <33CA3566.47C1@blackmagic.tait.co.nz>
Organization: Tait Electronics NZ
MIME-version: 1.0
Content-type: text/plain; charset=us-ascii
Content-transfer-encoding: 7bit
References: <33C8A01A DOT E8E AT mcs DOT nl> <6ZxdyDA4qMyzEwjH AT talula DOT demon DOT co DOT uk>
Precedence: bulk

Shawn Hargreaves wrote:
> Maybe, but remember that many djgpp programs are run under environments
> like Win95 that are not 'decent' memory managers :-) And depending what
> you do in your handler, it's unlikely that all the code and memory it
> can ever access will be touched every time it is called. Given the
> consequences of reentrant calls to the DOS disk I/O functions (in the
> worst case, a clobbered filesystem), it just isn't worth taking the
> risk.

And it's not difficult to do, especially using (Alaric William's idea)
the __attribute__((section)) trick.  Unfortunately, this requires a
patched gcc (next version (2.8?) will have the patch) and a fixed gas
from binutils 2.8.1 (I can supply the patched gas.exe and Alaric has the
patched gcc on his web page).

And another reason for locking your interrupt memory: cwsdpmi won't
allow the program to run if it doesn't lock the memory (cwsdpmi detects
accesses to unlocked memory in interrupt handlers and aborts the
program).

Bill
-- 
Leave others their otherness.