From: sandmann AT clio DOT rice DOT edu (Charles Sandmann)
Message-Id: <9608081324.AA09546@clio.rice.edu>
Subject: Re: window sti/cli bug
To: bill_currie AT MAIL DOT TAIT DOT CO DOT NZ (Bill Currie)
Date: Thu, 8 Aug 1996 08:24:32 -0600 (CDT)
Cc: djgpp-workers AT delorie DOT com
In-Reply-To: <s20a17f1.030@MAIL.TAIT.CO.NZ> from "Bill Currie" at Aug 8, 96 04:38:56 pm
Content-Type: text

> the iret (even though the dpmi spec says this should be done).  Here is a list of
> all occurences of the iret instruction (including data for _go32...), to be

> debug/common/dbgcom.c:171:Lc31a:	iret								\n\
Sofware int, OK.
> libc/ansi/setjmp/longjmp.s:83:	iret			/* actually jump to new cs:eip
Simulated int, OK.
> libc/go32/exceptn.s:284:	iret
Already has sti.
> libc/go32/exceptn.s:359:	iret
Timer speedup tic ignore needs sti, but this code is currently never called.
> libc/go32/exceptn.s:378:	iret
> libc/go32/exceptn.s:381:	iret
> libc/go32/exceptn.s:388:	iret
These are software interrupts.
> libc/go32/gopint.c:75:/* 53 */ 0xcf,				/*     iret                    
> libc/go32/gopint.c:81:/* 5A */ 0xcf,				/*     iret                    
> libc/go32/gormcb.c:71:         0xcf				    	/* iret                
> libc/go32/gormcb.c:78:         0xcf					/* iret                   
Probably should be fixed, but it's got other bugs in it too.  Needs to be
reentrant.

Except for the notoriously buggy gopint/gormcb calls, the libc looks OK
to me for anything you were likely to call.