X-Authentication-Warning: delorie.com: mail set sender to djgpp-workers-bounces using -f
X-Recipient: djgpp-workers AT delorie DOT com
Message-ID: <4DDD1D90.4000000@iki.fi>
Date: Wed, 25 May 2011 18:17:36 +0300
From: Andris Pavenis <andris DOT pavenis AT iki DOT fi>
User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.2.17) Gecko/20110428 Fedora/3.1.10-1.fc14 Lightning/1.0b2 Thunderbird/3.1.10
MIME-Version: 1.0
To: djgpp-workers AT delorie DOT com
Subject: Re: GCC-4.6.0
References: <4DDB35D2 DOT 7030600 AT iki DOT fi>
In-Reply-To: <4DDB35D2.7030600@iki.fi>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com

On 05/24/2011 07:36 AM, Andris Pavenis wrote:
> The first test build of gcc-4.6.0 for DJGPP is available at
>
> http://ap1.pp.fi/djgpp/gcc/4.6.0/
>
> There is also Linux to DJGPP cross-compiler RPMs available
> (should hopefully work on recent RPM based distributions)
>
> It took some efforts to get also libquadmath to be built
> at all and 'REAL (KIND=16)' support enabled in GNU fortran.
>
> However REAL(16) support is broken even worse rather
> than for Linux x86_64.
> - REAL(16) operations seems to work OK
> - Assigning constants or using REAL(16) constants in
> expressions does not (in Linux x86_64 only second
> of them does not work: I tested on Fedora 15 development
> version). The problem is with FORTRAN Quad precision
> constants (eg 1.2345678912345Q1 misinterpretted as
> a single precision constant). I have not yet had time
> submitted bug report though.
>
> Should I leave it on at all for DJGPP?

Well: see

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48720

I should have used E0_16 instead of Q0. My FORTRAN knowledge
has got a bit rusty as I have not followed the development
of FORTRAN standard. Anyway Qnn for quad float constants is
going to work for the following 4.6.X versions according to
the bug report mentioned above.

So I changed to use of <Enn_16> form for literals. After that
the program I used for testing (after adapting to use quad
precision) seems to work reasonably well

Andris