make install is pretty broken
Mark Kettenis
kettenis@wins.uva.nl
Sat Jul 29 16:26:00 GMT 2000
Date: Sat, 29 Jul 2000 12:12:58 -0700
From: "H . J . Lu" <hjl@lucon.org>
On Sat, Jul 29, 2000 at 05:47:51PM +0200, Mark Kettenis wrote:
> If force-install = no, make install is broken, see the attached bits
> from an install log. The problems are probably caused by the
> following patch:
>
> 2000-03-17 Cristian Gafton <gafton@redhat.com>
>
> * elf/Makefile ($(rtld-version-installed-name)): Delay replacing
> existing rtld-installed-name.
> ($(rtld-installed-name)): Request libc-$(version) be installed too.
> replace existing rtld-installed-name using $(symbolic-link-prog).
I have asked it a few times before. But I got zero positive responses:
http://sources.redhat.com/ml/libc-alpha/2000-07/msg00047.html
http://sources.redhat.com/ml/libc-alpha/2000-07/msg00048.html
Well, you've got one respsone now :-).
We can only guess at what problem Cristian's patch tries to solve, but
I guess it's overwriting glibc-x.y.z with glibc-x.y.z when there's
been a subtle change in the interface between libc.so and ld.so.
I think we do agree that Cristian's patch is (obviously) wrong. We
have a target `$(inst_slibdir)/$(rtld-version-installed-name)' that
only produces `$(inst_slibdir)/$(rtld-version-installed-name).new'.
Your patch seems to fix that, but partly reverts Cristian's patch. I
have no objection to that. The problem only occurs during
development. People installing a development version of glibc should
now how to fix their system when it breaks in this situation. And
Cristian's free to submit a new, correct, patch to solve his problems.
Mark
More information about the Libc-hacker
mailing list