From: <ams AT ludd DOT luth DOT se>
Message-Id: <200306192256.h5JMu1px028229@speedy.ludd.luth.se>
Subject: Re: Bugs in unassmbl.c
In-Reply-To: <Pine.OSF.4.51.0306171447140.29318@sirppi.helsinki.fi>
 "from Esa A E Peuha at Jun 17, 2003 03:19:44 pm"
To: djgpp-workers AT delorie DOT com
Date: Fri, 20 Jun 2003 00:56:01 +0200 (CEST)
X-Mailer: ELM [version 2.4ME+ PL78 (25)]
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Type: text/plain; charset=US-ASCII
X-MailScanner: Found to be clean
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

According to Esa A E Peuha:
> There are two bugs in unassmbl.c (both edebug and fsdb versions): arpl
> has wrong second operand, and frstor is in wrong place.  Are these OK to
> commit?
> 
> Index: src/debug/edebug/unassmbl.c
> ===================================================================
> RCS file: /cvs/djgpp/djgpp/src/debug/edebug/unassmbl.c,v
> retrieving revision 1.5
> diff -u -r1.5 unassmbl.c
> --- src/debug/edebug/unassmbl.c	1 Jun 2002 09:27:41 -0000	1.5
> +++ src/debug/edebug/unassmbl.c	17 Jun 2003 11:49:20 -0000
> @@ -107,7 +107,7 @@
>    "pop %eax", "pop %ecx", "pop %edx", "pop %ebx",
>    "pop %esp", "pop %ebp", "pop %esi", "pop %edi",
>  /* 6 */
> -  "pusha", "popa", "bound %Gv,%Ma", "arpl %Ew,%Rw",
> +  "pusha", "popa", "bound %Gv,%Ma", "arpl %Ew,%Gw",
>    "%pf", "%pg", "%so", "%sa",
>    "push %I-vv", "imul %Gv=%Ev*%I-vv", "push %I-vb", "imul %Gv=%Ev*%I-vb",
>    "insb %Yb,dx", "ins%ew %Yv,dx", "outsb dx,%Xb", "outs%ew dx,%Xv",

This seems to do the right thing. I tried it and ``asm (".byte 0x63\n\t
.byte 0");'' went from "arpl [eax], [eax]" (which can't be true) to
"arpl [eax], ax" (which is correct AFAICT).

> @@ -281,11 +281,11 @@
>  /*16*/ "fiadd", "fimul", "ficomw", "ficompw",
>         "fisub", "fisubr", "fidiv", "fidivr",
>  /*24*/ "fild", 0, "fist", "fistp",
> -       "frstor", "fldt", 0, "fstpt",
> +       0, "fldt", 0, "fstpt",
>  /*32*/ "faddq", "fmulq", "fcomq", "fcompq",
>         "fsubq", "fsubrq", "fdivq", "fdivrq",
>  /*40*/ "fldq", 0, "fstq", "fstpq",
> -       0, 0, "fsave", "fstsww",
> +       "frstor", 0, "fsave", "fstsww",
>  /*48*/ "fiaddw", "fimulw", "ficomw", "ficompw",
>         "fisubw", "fisubrw", "fidivw", "fidivr",
>  /*56*/ "fildw", 0, "fistw", "fistpw",

This seems right too (after the patch i. e.), according to the Intel
manual, although it's confusing with 16, 24, 32... matching 0x8a,
0x8b, 0x8c.

(Perhaps "00... 08... 16..." should be changed to "00, opcodes
0x88... 08, opcodes 0x89... 16, opcodes 0x8a..."?) 


Right,

						MartinS