Date: Thu, 9 Apr 1998 18:42:59 -0700 (PDT)
Message-Id: <199804100142.SAA14921@adit.ap.net>
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
To: Kbwms <Kbwms AT aol DOT com>, bug-gcc AT gnu DOT com, djgpp AT delorie DOT com
From: Nate Eldredge <eldredge AT ap DOT net>
Subject: Re: Problem with Optimizer in GCC
Precedence: bulk

At 02:39  4/7/1998 EDT, Kbwms wrote:
>The following problem pertains to MSDOS.
>
>When compiling enquire.c using optimizer level three, serious
>problems occur.  For example, enquire reports that the number
>of bits in the significand of a float is 64.  It says exactly
>the same thing for type double.

Did you notice this comment in enquire.c?

    Don't use any optimisation flags: the program may not work if you do.
    Though "while (a+1.0-a-1.0 == 0.0)" may look like "while(1)" to an
    optimiser, to a floating-point unit there's a world of difference.

>Several other errors occurred during testing as shown in the
>listing which I have included in enquire.zip (attached).  It
>is especially troubling that an execution of the program will
>often leave the FP processor in an error state.  The result
>is a status dump which I have also included in the zip file.

This is because DJGPP does not reset the FPU on SIGFPE, leaving it filled
with garbage to crash the next thing to run.  I seem to recall there being a
good reason for this; however, I've forgotten what it was. Can somebody
remind me?

If there isn't a good reason not to, it's a simple fix: Just add `fninit',
or whatever it is, to `_exit' in crt0.S.

>The attached zip file, enquire.zip, contains
>
>    1. enquire.c

I didn't get this attachment, although that's probably a good thing (large
binary posts to comp.os.msdos.djgpp are discouraged). I assume you're using
the same enquire.c that comes with GCC? (I use 2.7.2.1.)

Nate Eldredge
eldredge AT ap DOT net