Re: [3.2 PATCH] install-no-fixedincludes repaired

[Christian Cornelssen <ccorn at cs dot tu-berlin dot de>]

Hi,

On Sat, 23 Nov 2002, Nathanael Nerode wrote:

> Christian, this patch:
>
> http://gcc.gnu.org/ml/gcc-patches/2002-10/msg00418.html
>
> although it works, is bad practice.
>
> Makefiles should *never* care about order of dependencies.
>
> I don't see a sufficient way to fix this correctly for 3.2, but
> I'm going to try to get it right for 3.4.  It probably involves breaking
> out some items into their own targets.

Good notice. However, many install* targets of e.g. `gcc/Makefile.in'
(as of gcc-3.2), including the commonly used ones, seem to be
susceptible to dependence ordering.  That is, if you already have
redesigned the common install* targets, just extending your approach
to `install-no-fixedincludes' should fix the problem. :-)

But as a consequence of what you have pointed out, I do not expect
my install-no-fixedincludes patch to get outside the 3.2 branch.
It's better than nothing and not worse than the other install*
rules, but in general it's not appropriate, I agree.

Regards,

Christian Cornelssen