Public SourceMage Discussion List

[Eric Sandall <eric AT sandall.us>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Thomas Orgis wrote:
> Hi,
> 
> glibc-2.11 has been released recently and you might be tempted to update 
> our spell. Well, .. oh ... fricked: Ladislav did it already. Hm. Uhm. Well. 
> What I was about to say:
> 
> 1. I am working on updating the SGML Alpha port. That means adding the 
> spin-off glibc-ports for non-x86 architecture and an assload of patches, by 
> way of the gentoo patchset, to battle against the ignorance of glibc 
> developers towards such unimportant architectures like ARM (which is not 
> officially an arch supported/maintained by glibc upstream anymore!). 
> 
> 2. glibc-ports people need to catch up with funny stuff in latest glibc to 
> make a matching release of glibc-ports-2.11 . They are not there yet, as I 
> see from reading the libc-ports list.
> 
> 3. Likewise, as even glibc-ports won't manage to provide working Alpha 
> support in release version 2.11, I depend on the gentoo patchset (which 
> largely is the debian patchset, as alpha fixes are concerned) to get a 
> working glibc at all. The friendly gentoo guy said he is going to check out 
> glibc 2.11 in the near future, but, well, it's not been that long that 
> 2.10.1 works.
> 
> 4. One might indeed consider ditching vanilla glibc as upstream. Eglibc is 
> getting some time in the sunshine now that debian switches to it... might 
> be reasonable to use that as a buffer between our users and the not always 
> too human-friendly upstream actions. I mean, I am a bit frightened by the 
> kind of patches that are needed to transform vanilla glibc into something 
> barely usable. And indeed the lack of stable release series... like, with 
> bugfixes and minimal other hacking, is not nice for something that 
> essential.
> 
> 
> Anyhow, I popped in to tell that I am close to getting our alpha port 
> working again with glibc-2.10.1 ... and now see that that is already p0wn3d 
> by 2.11 entering the grimoire. The kind of discussions on libc-ports 
> suggests that it is non-trivial to follow the changes in glibc... Now we 
> could have a debate over whether such esoteric non-x86 setups should hinder 
> the progress of the important SMGL world... or figure out a way to make it 
> work nicely together.
> Personally, I would think it fair to wait for glibc-ports to appear before 
> upgrading to a new glibc version. glibc + glibc-ports is what earlier 
> versions of glibc represented.
> 
> Since there already is alpha-specific code needed in the spell scripts (as 
> for x86-64, too); it might be not too bad to use version 2.10.1 for that 
> arch and 2.11 on others. What bothers me is that that important spell 
> already is cluttered, with NTPL and non-NPTL, CVS version (do we really 
> need that?)... then various if branches for SMGL_COMPAT_ARCH ...
> 
> It would already help to drop non-NPTL and CVS ... perhaps then I can 
> better stand having only the multi-version switch based on architecture. 
> So, who uses those two options? Do we support kernel 2.4 (for non-NPTL)?
> 
> Flames, anyone? Man, it's late here... Perhaps I shouldn't even send this 
> babbling, but when I already spent all the time deep into the night to 
> write it... trying to get that fricking bleeding edge glibc crap to work... 
> gah, I'm drifting off again...
> 
> 
> Alrighty then,

I'd say just add a specific version for your architecture and wrap it in
 DETAILS in an architecture check.

- -sandalle

- --
Eric Sandall                     |  Source Mage GNU/Linux Developer
eric AT sandall.us PGP: 0xA8EFDD61  |  http://www.sourcemage.org/
http://eric.sandall.us/          |  http://counter.li.org/  #196285
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAkr43OYACgkQHXt9dKjv3WHH5gCgnaJNg78G7NP2jfTF+xvrQfyi
2UoAoJypcOAPXwbgxTq4dbdNY2tH5Nv9
=fKGN
-----END PGP SIGNATURE-----