bug-gnulib
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[bug-gnulib] Re: base64


From: Simon Josefsson
Subject: [bug-gnulib] Re: base64
Date: Sat, 27 Nov 2004 17:07:35 +0100
User-agent: Gnus/5.110003 (No Gnus v0.3) Emacs/21.3.50 (gnu/linux)

Stepan Kasal <address@hidden> writes:

> Hi,
>
> On Sat, Nov 27, 2004 at 01:18:46AM -0800, Paul Eggert wrote:
>> >> I guesss this table assumes ASCII.  Do you want your code to be
>> >> portable to EBCDIC hosts?
> [...]
>> A bit crude, but the only other ways that spring to mind involve using
>> Autoconf.
>
> I have probably missed something, but what would be wrong if autoconf
> were used to detect whether the host uses EBCDIC or ASCII?
>
> gnulib modules usually deliver with snippets of autoconf code, right?

I don't know the official position, but I prefer to keep the autoconf
magic to a minimum.

To me, code should ideally build without any autoconfery magic on
compliant C89/C99/POSIX platforms.  If it was possible to get rid of
all traces of autoconfery in the source code (like getting rid of
'#include "config.h"') I would want that.  Autoconfery stuff is to me
something you are forced to use because of broken platforms.

On the other hand, once you are using it, it can be argued that you
should take full advantage of it.

Besides, if the code was written in good style, it would not need
autoconf to discover ASCII vs EBCDIC, and would also work with new
charsets in the future.  I think that is a better solution, if we know
how to do it.

This point of view also helps when someone wants to re-compile GNU
using a safe dialect of C (like Cyclone), or experiment with the code
in general.  Just make your tools support C99 and POSIX, and the code
should compile.  Fixing others' broken autoconf tests is not that
fun..

Thanks.





reply via email to

[Prev in Thread] Current Thread [Next in Thread]