Santiago Vila <[EMAIL PROTECTED]> writes:
> On Mon, 10 Nov 2003, Eike Sauer wrote:
>
> > What about letting Robert build and upload (if ftp-masters agree)
> > his package, *if* he puts it in experimental, uses a description
> > that contains a warning about the experimental status of the
> > package in a prominent place, and not calling it "linux", but...
>
> linux-2.4.22 please, but just the binary-package, it should be ok to call
> "linux" the source package.
>
> This way we could put an end to the objection that it may not be
> upgraded safely.
>
>
> As far as unstable vs experimental is concerned, I think one of the
> goals for this package is to have a common source package for the
> autobuilders. Since the autobuilders do not build experimental, it
> would not make any sense at all to upload it for experimental.
It will _only_ built on i386. Further more even with changes it will
only work on a very few archs so far.
And the package is 27MB too big since it should depend on the vanilla
linux sources instead of including them.
> If Robert is such an incompetent developer as some people say and the
> package does not build on the 11 different architectures, then the
> package will not propagate to testing and the world will be safe from
> the disaster.
A package will enter testing if it builds on all (or more) arch it was
previously build. Someone would have to file a FTBFS bug before it
becomes a candidate.
> OTHO, if he manages to create a package which compiles on every
> architecture and produces an *usable* kernel on every of them, why
> should not be the package allowed to exist in testing?
>
> And if there is some architecture on which the kernel produced is not
> usable, would this not be a good reason for a "severity: serious" bug,
> which would again save the world from the disaster?
He has a lot of work before him before its useable. Personally I think
the time is better spend on merging the existing images source
together with the old build technique then to start from scratch.
MfG
Goswin
