Xref: utzoo comp.windows.ms.programmer:3094 comp.lang.pascal:7148
Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!usc!orion.oac.uci.edu!ucivax!milne
From: milne@ics.uci.edu (Alastair Milne)
Newsgroups: comp.windows.ms.programmer,comp.lang.pascal
Subject: Re: Question on UAE's when running TPW
Message-ID: <285828DC.6958@ics.uci.edu>
Date: 14 Jun 91 02:24:27 GMT
References: <91157.094647F0O@psuvm.psu.edu> <1991Jun12.031916.1760@chshost.in-berlin.de>
Distribution: na
Organization: UC Irvine Department of ICS
Lines: 40

In <1991Jun12.031916.1760@chshost.in-berlin.de> root@chshost.in-berlin.de (Thomas Liening) writes:

>F0O@psuvm.psu.edu writes:
>>I was wondering if your other programs are protected from
>>these errors or not.  I wouldn't like to have to exit windows and restart
>>every time I would encounter a UAE, but maybe this is the best way to
>>work?
>I don't think that other programs are protected from UAEs in one
>application, since an UAE can easily crash the *entire* Windows system
>if it's serious enough.

    Somebody from Microsoft posted a note the other day saying that
    Windows only uses 1 LDT (Local Descriptor Table) for all tasks.
    For those who may be unfamiliar with the 286 and 386 arrangements,
    the Descriptor Tables are, among other things, how the chip translates
    each task's memory references into actual physical memory, segregated
    from each other -- that's what the word "protected" means in "protected
    mode".  Each task is supposed to have its own LDT -- in fact, that's 
    essentially the {2,3}86's definition of a task.  If Windows is not
    creating a new LDT for each task, then they have no protection
    against each other that I am aware of.

>>   I've heard other people say when you're running other windows programs
>>and get a UAE, you should exit ASAP to avoid any possible further damage.

    That's what the Cobb Group people say in Inside MS Windows.  Sounds
    terrible to me, but I've certainly had DOS corrupted from some 
    of these.

    By the way, I should point out that Irrecoverable Errors (I'll swear
    there's no such word as "unrecoverable") are not the only way for 
    Windows to die.  I've had operations proceeding when suddenly Windows
    is gone and I'm staring at a blank DOS screen.  (To be fair, I *think*
    these happen when I'm trying to use DOS shell sessions within 
    windows, and they may be trying to install interrupt handlers within 
    their virtual machines.)  However it happens, though, the system
    is usually fit for nothing at this point, and should be rebooted.


    Alastair Milne