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

Lluís Batlle i Rossell viriketo at gmail.com
Sat Oct 30 21:03:10 CEST 2010


On Sat, Oct 30, 2010 at 10:30:46AM -0400, roconnor at theorem.ca wrote:
> On Sat, 30 Oct 2010, Ludovic Courtès wrote:
> 
> >OTOH, we don’t want to make bootstrapping too complex either, so there’s
> >surely a trade-off here, and it probably needs to be considered on a
> >case-by-case basis.  For this particular case, avoiding the upgrade
> >would have been costly, I guess.
> 
> In this case, The cost of avoiding the upgrade is needing to compile
> gcc 4.4 twice.  Once to be used to build the new glibc and a second
> time to built with the new glibc.
> 
> I personaly don't find this to be a big burden and it would be what
> I would do; however I'm not nearly as experienced as Eelco, so I
> trust his judgement.

I invite you to prepare gcc expression to allow that :)

I mean that we never tried having such a 'middle-gcc' in the bootstrap. So it
may be not that simple as adding a gcc build in the stdenv bootstrap, as
we may have to build it somehow specially.

For Peter, I think that if he can afford using a different glibc in nixpkgs, he
can also afford using older bootstrap-tools. If his system cannot run glib2.12,
it does not matter if it's in bootstrap-tools or it is built by any old
bootstrap-tools. For what I understand until now, I think he can afford that.
That reduces the most immediate problem to "whether to use disk space with
new bootstrap tools in the nix subversion repository or not", and we should not
think that the new bootstrap-tools will disallow Peter from running nixpkgs.
Correct me if I am wrong.

What if we add the gcc stage in *the next* stdenv-updates, and not in this
merge to trunk we are trying to do since more than one month, maybe?

Otherwise, if we delay it much, some may start using stdenv-updates instead of
trunk. :)

Regards,
Lluís.



More information about the nix-dev mailing list