[ On Friday, July 28, 2000 at 16:35:57 (+0200), Bruno Haible wrote: ]
> Subject: Re: need to define _ISOC99_SOURCE
>
> I'd deprecate it. It solved a problem with SunOS 4.0.x long ago. We
> shouldn't make people think that -traditional is good for anything,
> when in fact more and more sources cannot be compiled in K&R mode.
True, however there should be no reason, and not even an implication,
that Autoconf can't be used to make existing K&R code more portable.
I.e. please don't force those of us who maintain big chunks of K&R code
to convert it entirely -- that's neither practical nor necessary to keep
it working just fine. There are (still) a few bugs in modern standards
that prevent perfectly fine K&R constructions from working the same as
they always did (so much for standardisation!), but those can be worked
around in ways that keep the code compatible with K&R compilers and
those are the only changes I'm willing to make. Fancy shiny new
compilers are not always suitable even if they can be made to work on
the platforms in question (or cross-build for it). Good software
configuration management practices show that the tools are part of the
software configuration and while it may be enormously beneficial to make
code portable to other platforms (and thus change some of the build
processes), changing the compiler unnecessarily in a production
environment is a very bad thing to do without full QA testing!
BTW, the only reason lots of recent code can't compile in K&R mode is
because its authors are either lazy, or ignorant (of C's heritage), or
both! ;-)
C != C++
--
Greg A. Woods
+1 416 218-0098 VE3TCP <[EMAIL PROTECTED]> <robohack!woods>
Planix, Inc. <[EMAIL PROTECTED]>; Secrets of the Weird <[EMAIL PROTECTED]>
