Re: possible compromise for ITP: linux?
Eike Sauer <email@example.com> writes:
> Andrew Suffield schrieb:
> > He doesn't need to, he can be slapped down.
> "Keine Gewalt!" ("No violence!")
> > We don't ignore minor issues just because there are major ones.
> So let's hope Robert can cope with minor issues
> and only talk about the big ones for now.
> > - this packages adds nothing, and would occupy a fair chunk of space
> > in the archive.
> I don't know how short Debian is of space.
> How large would Robert's packages be?
> There already are several packages with complete
> kernel sources which take as much place as his package
> would, right?
> So is this really too large to let him test his idea?
Lets look at sid and just sources:
mrvn@dual:/mnt/debian/pool/main% find -type f -name "*kernel*gz" | xargs du -h --total | grep total
mrvn@dual:/mnt/debian/pool/main% find -type f -name "*kernel*gz" | xargs du -m | sort -nr | head
As you can see (apart from hppa and ia64, what are you guys doing
there?) there is only one kernel source for each version.
All the kernel-image packages build-depend on the one kernel source
package and various kernel patch packages and are only a few K big
-rw-rw-r-- 1 mrvn mrvn 673 Jun 24 23:47 kernel-image-2.4.20-amiga_2.4.20-5.dsc
-rw-rw-r-- 1 mrvn mrvn 4.8K Jun 24 23:47 kernel-image-2.4.20-amiga_2.4.20-5.tar.gz
There is no reason to duplicate the kernel source even if the images
are to be produced differently.
> > - this package cannot be safely upgraded (without forcing a reboot).
> > The latter prohibits it from being in a Debian release.
The same can be said for the current images.
> So it doesn't stop it from entering experimental.
> But I'm not sure if a package is worth a try if
> it cannot possibly make it in a release.
Duplicating the vanilla source in the archive is a very bad
idea. Thats an extra 40 MB for every 2.6 kernel to be.