[Nix-dev] Re: Please update of stdenv-linux tarballs

[Lluís Batlle i Rossell]

On Tue, Nov 02, 2010 at 11:53:23AM +0100, Peter Simons wrote:
> Hi Lluís,
> 
>  >> In my humble opinion, a sensible guideline would be that bootstrap tools
>  >> are not supposed to be modified unless they have to be. [...] Now, I
>  >> understand Lluís needs the latest binutils in bootstrap tools. That is a
>  >> perfectly legitimate interest, and I am all for it. I wonder about the
>  >> update to glibc 2.12.1, though. Is that really necessary to support
>  >> loongson2f?
>  >
>  > No, the glibc 2.12.1 was not necessary.
> 
> well, then this is one heck of an unnecessary update. By now, it's too late
> to do anything about it anyway, so I guess we just move forward and deal
> with any consequences that might arise.

I still don't know what problem can come from having a newer glibc in the
bootstrap-tools. As I said, I think what matters for your kind of troubles are
the headers of the linux version used. I think newer versions will simply work
better. Older glibc have bugs and exploits, as we have seen these days.

> For the future, however, I would like to suggest that major updates to
> bootstrap tools are discussed on this list beforehand. Different people have
> different interests and concerns, and because of that it's wise to involve
> all interested parties in those kind of decisions, so that everyone has the
> chance to communicate their interests.

Making the bootstrap-tools is making a single derivation. It's not about getting
the old bootstrap-tools and changing some files in it. And the files in it have
to be built coping all with each other. We have the bootstrap-tools in a single
package - if we have to update the package because one piece works bad, we can
update all. Using newer versions of the software should only be an improvement.
Requiring a newer kernel, though, may not "be an improvement". If you don't agree,
please give details on that. And keep in mind the final nixpkgs binaries don't
depend only on the bootstrap-tools, but also on the rest of stdenv components.

Of course I'd like to involve all the parties interested, but for that to be
useful, we need enough information on the problem we deal with.

So, please, I ask again:
- What linux kernel runs CentOS? At least the version, please.
- What makes you think the glibc version is the important difference in
  determining what runs and what not? Couldn't be it the kernel headers?
- What have you been doing, in detail, to get nixpkgs built on your CentOS?

We may use this information for the next merge, instead of this. But please
provide either the information, or patches solving the problems.