To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
Cc: Andris Pavenis <pavenis AT lanet DOT lv>,
        "Gurunandan R. Bhat" <grbhat AT unigoa DOT ernet DOT in>,
        Alexey Yakovlev <jack AT catalysis DOT nsk DOT su>, djgpp-workers AT delorie DOT com,
        fortran AT gnu DOT org
Subject: Re: Inconsistencies between g77 v0.5.23 and v0.5.19
References: <Pine DOT SUN DOT 3 DOT 91 DOT 980712192223 DOT 27114O-100000 AT is>
From: Dave Love <d DOT love AT dl DOT ac DOT uk>
Date: 12 Jul 1998 19:19:27 +0100
In-Reply-To: Eli Zaretskii's message of "Sun, 12 Jul 1998 19:22:48 +0300 (IDT)"
Message-ID: <rzqpvfb2aq8.fsf@djlvig.dl.ac.uk>
Lines: 30
Precedence: bulk

>>>>> "Eli" == Eli Zaretskii <eliz AT is DOT elta DOT co DOT il> writes:

 Eli> DJGPP shows that `unix' can be defined by a compiler which
 Eli> targets other systems.

Sure, and there are already special cases for Windows32 stuff, but
we're interested in a characteristic of Unix (POSIX, I guess).

 Eli> In general, it is best to test for a feature rather than for a
 Eli> name of an operating system.

Sure, where possible.  (Perhaps we should now case-analyse configure's
target name, but that wasn't easy previously and may still not be in
gcc 2.8.)

 Eli> Perhaps that ifdef could be rewritten so that it tests for the
 Eli> actual functionality?

I don't think so.  The behaviour for cygwin32, at least, depends on
how you mount the file system, not on the library intrinsically.  Also
we need to be able to (criss-)cross-build, which means we can't rely
on running compiled test cases.

 Eli> Since the patch is for configure.in, the DJGPP version *must* be
 Eli> built by running the configure script.

But not necessarily natively under DJGPP.  Also gcc still contains
configur.bat, which doesn't support f77.  I have asked before about
g77 support for DJGPP without luck and was never motivated to
investigate myself.