Xref: news2.mv.net comp.os.msdos.djgpp:4378
From: Thomas Demmer <demmer AT LSTM DOT Ruhr-UNI-Bochum DOT De>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: buggy libm.a
Date: Tue, 28 May 1996 09:06:48 +0100
Organization: Lehrstuhl fuer Stroemungsmechanik
Lines: 47
Message-ID: <31AAB418.4487@LSTM.Ruhr-UNI-Bochum.De>
References: <199605241956 DOT AA09563 AT sun17 DOT thp DOT Uni-Koeln DOT DE>
NNTP-Posting-Host: bvb.lstm.ruhr-uni-bochum.de
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: Wolfram Brenig <brenig AT thp DOT Uni-Koeln DOT DE>
DJ-Gateway: from newsgroup comp.os.msdos.djgpp

Wolfram Brenig wrote:
> 
> Ok. you DJGPP gurus please give this one a try:
> 
> On May 15 and 19 I posted questions regarding problems I encountered
> with SIGFPE's which I get with V2.x and lengthy algebraic exprs. compiled
> using the -On switch. The only more serious reply I got was from Thomas
> Demmer <demmer AT lstm DOT ruhr-uni-bochum DOT de> telling me that there is a *bug*
> in libm.a ?!
> C'mon you people out there, anybody else that would like to confirm that
> the basic math library of V2.x is *really* buggy ?!?!? (Like: '2+2=-17.5
> and lets hope that this is the only one' ...)
> 
> Any help is most appreciated. I am not subscribing to the djgpp mailing
> list so please mail your comments to: brenig AT thp DOT uni-koeln DOT de
> 
Not being a guru I respond anyway. The only known bugs in 
libm.a I know of are caused by a missing pop on the floating
point stack after a FSCALE command. Affected functions are ldexp()
and exp(), maybe some descendants of them. I grep'ed the sources for
FSCALE and added the missing pop. The fixed lib can be found on
ftp://ftp.lstm.ruhr-uni-bochum.de/pub/djgpp/libm.zip
as I didn't know any better place to put them to. 

If you're getting the above mentioned results, this would rather be an error
in gcc itself.

Right now I'm thinking about a program to "certify" the libm, checking most of
the functions. But I can only cross-check that on an IBM with IBM's cc, which is
_known_ to be pretty buggy itself.

-- 
Ciao
Tom



*************************************************************
* Thomas Demmer                                             *
* Lehrstuhl fuer Stroemungsmechanik                         *
* Ruhr-Uni-Bochum                                           *
* Universitaetsstr. 150                                     *
* D-44780  Bochum                                           *
* Tel: +49 234 700 6434                                     *
* Fax: +49 234 709 4162                                     *
* http://www.lstm.ruhr-uni-bochum.de/~demmer                *
*************************************************************