Wow, support arrived ;) j/k..

On Tue, Feb 04, 2003 at 03:41:16PM -0000, Stefan Jones wrote:
> Jozsa Kristof said:
> > facts:
> > - oracle 9i wont work with glibc 2.3.1 (it worked with glibc 2.3) - one
> > oracle employee confirmed that on the forums in an answer to someone -
> > portage has no more glibc 2.3 ebuilds, just 2.3.1
> 
> I was unaware of any major changes from 2.3 to 2.3.1; all that was changed
> were bug fixes to allow applications to run with small stack sizes. 2.3
> can be put back.

Ok, that's definetely good news.

> > questions:
> > - how do i do a proper downgrade to glibc 2.2?
> 
> Well you would have to recompile everything, this is a bootstrap situation.

I drop this one.. I prefer going back to the oracle-compatible glibc 2.3.
I've recompiled and replaced the whole libc successfully meanwhile back to
2.3.1 (from 2.2) using the bootcd, chroot and some mount -o bind tricking..
fun fun fun ;)

> > - or where do i get a glibc 2.3 ebuild from to get back our oracle
> > working? - why gentoo maintainers force people and leave no choice, as
> > removing essential packages like glibc 2.3 from the portage tree?
> 
> Make one yourself, copy the ebuild to glibc-2.3.ebuild, and quickly scan
> the file for any 2.3.1 -> 2.3 changes needing to be done. There have been
> no changes between 2.3.1 and 2.3 of any note in the ebuild. Some / All
> patches may need to be removed though.

Right, I'll fight with this myself then. In spite of this, it'd be _really_
helpful not to remove ebuilds from the portage tree to avoid such
situations.. at least for important packages.

> > - if they do so, will really this pile-o-shit make itself the stable
> > 1.4 released within days, and gentoo people are disabled to run oracle
> > 9i anymore?
> 
> Well if you want to do things by the book, only use software sanctioned by
> Oracle (ie. old redhat distros) . It is not gentoo's or GNU/Linux's fault
> the Oracle ppl cannot keep up and release POSIX / glibc complient binaries
> or recompile for newer glibc's. (if this is false then post a glibc bug
> for not being POSIX )

Strictly watching it's a gnu fault that glibc 2.3.1 aint support the
LD_ASSUME_KERNEL option properly, while glibc 2.3 did that, and this wasn't
noted anywhere either. Again, such a change should be documented in proper
distributions as well, in spite of the fact that gentoo is meant for the
desktops mainly. (I believe Gentoo *is* one proper distrib)


Summing it up, whether gentoo is meant for what use mainly, if the upcoming
gentoo release's Oracle-compatibility depends on including the glibc 2.3
ebuild and adding a warning to the docs, I believe it's well worth these
heroic efforts.

Thanks for your notes and for your work on gentoo,

Christopher

ps. btw, Jakub Jelinek's glibc changelog notes that some work was
being done on LD_ASSUME_KERNEL in the glibc 2.3.1-12 version on early
december.. but I did not tested this version so cannot comment on this
compatibility issue any further.
-- 
.Digital.Yearning.for.Networked.Assassination.and.Xenocide

--
[EMAIL PROTECTED] mailing list

Reply via email to