From: pavenis AT lanet DOT lv
Message-ID: <B0000094300@stargate.astr.lu.lv>
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>, djgpp-workers AT delorie DOT com
Date: Mon, 12 Jul 1999 15:13:49 +0300
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Re: Operand-size suffixes
References: <B0000094278 AT stargate DOT astr DOT lu DOT lv>
In-reply-to: <Pine.SUN.3.91.990712131356.1868D-100000@is>
X-mailer: Pegasus Mail for Win32 (v3.11)
Reply-To: djgpp-workers AT delorie DOT com

On 12 Jul 99, at 13:23, Eli Zaretskii wrote:
 
> On Mon, 12 Jul 1999 pavenis AT lanet DOT lv wrote:
> 
> >      - some unnecessary operand length prefixes which are no more
> >        accepted by new binutils snapshots
> 
> Your previous message (the one from January) only mentioned FP 
> instructions that don't reference memory.  "fldcww  2(%esp)" does 
> reference memory, so I didn't change that.

FLDCW and FSTCW have predefined memory operand size (word 
only). It would be usefull not to force user to say it once more. 
At least not defining operand size explicitly works Ok with binutils-
2.9.1

> Is there any place where these changes in Binutils are described?  I 
> would like to avoid new problems with every new snapshot, if possible.

I didn't find it after some grepping docs. Therefore perhaps in 
sources only (gas/config/tc-i386.c ?)

Maybe it would be usefull to ask in binutils mailing list

> > 	- missing prototype for 
> > 		long double hypotl(long double x, long double y);
> >         in math.h
> 
> hypotl is not in the library, so no prototype is needed.  The source is 
> there, but it isn't compiled when the library is built (well, at least it 
> shouldn't be compiled).

At least when I run make from src it was attempted to build and I 
got a warning about missing prototype. Removed hypot.o and run 
make in that directory and I got:

C:\TMP\djgpp-990711\src\libc\ansi\math>make
gcc ... -c hypot.c

C:\TMP\djgpp-990711\src\libc\ansi\math>

So my suggestion is to rename hypot.c to hypotl.c

> We now have hypot.S (written by Eric Rudd) as the source of hypot
> functiopn.  I left hypot.c in the sources after some trivial changes
> to make it hypotl, but since it is largely untested and undocumented, I
> didn't add it to the makefile.  It is there for the time when C9X is
> out, which will most probably require long double functions like hypotl. 
> 
> > 	- one warning in src/debug/fsdb/fullscr.c where argument of
> > 	  wrong type  were given to sprintf()
> 
> Thanks, I corrected that already a day or two ago.
> 

Strange I updated source from CVS this morning

Andris