"Tim Robinson" <tim.at.gaat.freeserve.co.uk@invalid.com> wrote in message
news:bdcril$rq68o$1@ID-103400.news.dfncis.de...

Yes, you are right.
But talking was about "DOS programs".

--
George Pavlenko

CBFalconer wrote:
This would imply that Flat Real Mode doesn't work at all, and I can
assure you it does.

I don't know where the idea that modifying segregs resets the limits
came from. Does it say this in the Intel manuals? (I'm not as familiar
with the manuals as I should be) As nearly as I can determine, the only
way to alter limits is to jump to pmode, load segregs with a descriptor
containing the desired limit, and go back to rmode. If the limit was
4G-1, we're in Flat Real/Unreal mode. If something else does this, and
sets the limits back to 0FFFFh, then yes, you'd get a segment overrun
exception if you continued to try to use FRM. Bios calls might do this,
I can't imagine a dos call doing it. Herman's code installs the switch
to FRM as a handler for the segment overrun exception, so even this
eventuality is handled. Seems rock-solid to me!

As Tim says, OT for the win32 group... sorry.

Best,
Frank

"Tim Robinson" <tim.at.gaat.freeserve.co.uk@invalid.com> wrote:

Hence the big problem with wide cross-posting:

Newsgroups: alt.lang.asm,
comp.os.msdos.programmer,
comp.os.ms-windows.programmer.win32,
comp.programming,
microsoft.public.masm

In many of those newsgroups, the Windows assumption would not be so
natural.
--
- Tim Roberts, timr@probo.com
Providenza & Boekelheide, Inc.