Re: GCC 3.0.2 and BSD Make

[Zack Weinberg <zack at codesourcery dot com>]

On Fri, Nov 02, 2001 at 12:35:00AM +0100, Marc Espie wrote:
> 
> Well, if it weren't for H.J.Lu, I don't think anyone here would seriously 
> suggest gcc actually change to gnu-make only. This would be an utterly 
> bogus move in terms of robustness anyways.

On the contrary - I have in the past advocated this, and remain
convinced that it would be an unqualified improvement to robustness
and maintainability of the Makefiles.  Not so much because of all the
whizz-bang features, as because we could stop worrying about the
myriad inconsistent implementations of the features that we already
use.  VPATH, for instance.

There are very few extended-make features that I would like to be able
to use that we don't already; and again we can't use them not so much
because they aren't universally available, as because every
implementation did them differently and the semantics therefore cannot
be relied upon.  "include" is an obvious example.  As far as I know it
is universally implemented by modern Make implementations; with
consistent syntax even.  However, every implementation has different
rules for where the file is searched for and what happens when the
include-d file doesn't exist.  Therefore we can't use it.

The only genuine problem I know of with requiring GNU make is the
potential that it needs an ANSI C compiler to build with, producing a
chicken-and-egg problem on pre-standard systems.  However, this is
only a theoretical problem - on a brief skim of the code it certainly
*appears* to be K+R-safe.

I'm curious to know why you think requiring GNU make would be "an
utterly bogus move in terms of robustness". 

zw