Re: Bug#219582: ITP: linux -- Linux 2.4 kernel
Jamie Wilkinson wrote:
>Kernels install /boot/System.map-$version. There's a symlink from
>/boot/System.map to the current version.
And Robert's proposal currently results in the System.map-$version for
my current kernel vanishing, along with my modules.
>You are told you need to reboot after installing a kernel package.
No. You're currently told that you need to reboot after installing a
kernel package for the same kernel version as you're currently running.
This happens rarely.
>It looks to me like you're harping on a single issue which would have been
>encountered during the process of making this package, and based on this
>Robert is suddenly a second-class maintainer?
If someone is hoping to maintain a critical part of infrastructure, then I do
expect them to have some idea of what all the strange little files it provides
do and how messing things up could break them, yes. For instance, I wouldn't
even consider going anywhere near maintaining glibc, despite having messed
about with its guts quite considerably. I just don't understand the code
>If everyone in this project had to get the right answer first time, there
>would be a lot fewer maintainers and a lot fewer bugs in the BTS.
If you're uploading a package that otherwise already exists in the archive,
and your package fucks shit up that the other one doesn't, then you're
doing something wrong.
>>No, I'm saying that you're proposing to package a major piece of
>>infrastructure and give it a name that may attract users into installing
>>it, and the amount of thought and consideration that you seem to have
>>put in is insufficiently large for me to consider that it'll do anything
>>other than convince people that Debian kernel packagers are on crack.
>>Which would, again, be bad.
>I'd reiterate that you're implying Robert is going to make a half-arsed
>attempt and upload something he hasn't tested, but I've already said that.
I'm implying that because he's so far stated that various issues aren't real
problems. I don't think it'll be a half-arsed attempt, and I imagine that it
will be tested, but I think it's likely to prove of significantly lower
quality than the package that already exists. Frankly, I think that would be
true pretty much independent of who maintains it - Herbert has a good deal
of familiarity with the kernel, and many years of experience. The kernel
packages are decent. Uploading new packages with a name that's going to
attract people into installing them and which are likely to be of lower
quality is just insane.
>Besides, it's already evident that Debian maintainers (as a superset of
>kernel packagers) are on crack.
Most users don't find that packages they install break stuff.
Matthew Garrett | email@example.com