Control: severity -1 important Justification: essential package that works flawlessly for me
Hi,
First, I'd like to thank whoever has been maintaining kernel-package.
I depend on it, and the functionality enabled by this package is one
of the primary reasons I chose Debian over other alternatives--years
ago... Kernel-package is a killer feature! :-) Reply follows inline.
On Sun, Mar 24, 2019 at 03:31:27PM +0000, Ben Hutchings wrote:
> Package: kernel-package
> Version: 13.018+nmu1
> Severity: serious
>
> I discussed the state of kernel-package with Manoj and we agreed (see
> below) that it is not currently in a state suitable for release.
>
> Ben.
>
> ---
> From: Manoj Srivastava <manoj.srivastava.1...@gmail.com>
> Date: Thu, 7 Mar 2019 23:15:30 -0800
> Message-ID:
> <cabkpoev_od02cyamius4w7dc6ejz5bm5eefr0s4ta_w99jm...@mail.gmail.com>
> To: Masahiro Yamada <yamada.masah...@socionext.com>
> Cc: Ben Hutchings <b...@decadent.org.uk>, Linux Kbuild mailing list
> <linux-kbu...@vger.kernel.org>, Liz Zhang <liz...@microsoft.com>, Lili
> Deng <v-l...@microsoft.com>, Riku Voipio <riku.voi...@linaro.org>, Michal
> Marek <michal.l...@markovi.net>, Linux Kernel Mailing List
> <linux-ker...@vger.kernel.org>
> Subject: Re: [PATCH] kbuild: add workaround for Debian make-kpkg
>
>
> Hi,
>
> Well, if there are users, then I'll file the return bug, but I'll take a
> look to see how but rooted this is, and perhaps even try building personal
> imaged Deb packages again (I stopped doing that)
>
Count me as one of its users. I use it because it supports old-style
manually configured .config, and it lets one quickly and easily build
a Debian-packaged custom kernel like this:
make menuconfig
fakeroot make-kpkg binary-arch \
--initrd --append-to-version=.${date -I}
The new style kernel packaging is hard to learn how to use, and builds
take much longer for some reason (generation of more packages?).
> If should not go into buster in it's current state. It should be killed
> dead of it can't be brought back to a working state.
>
13.018+nmu1 on buster works like it always has for me--flawlessly. I
built upstream vanilla 4.19.31 two days ago.
> Manoj
>
> On Thu, Mar 7, 2019, 10:43 PM Masahiro Yamada <yamada.masah...@socionext.com>
> wrote:
>
> > On Fri, Mar 8, 2019 at 6:56 AM Ben Hutchings <b...@decadent.org.uk> wrote:
> > >
> > > On Wed, 2019-03-06 at 22:48 -0800, Manoj Srivastava wrote:
> > > > Hi,
> > > >
> > > > Does this have any users?
> >
> > Recently, I received a regression report.
> > So, yes. There are users.
> >
> >
> > > > I can take a stand at making it work, but I am
> > > > unsure of there are enough people interested in make-kpkg anymore to
> > make
> > > > it worthwhile.
> > > >
Please count me as a supremely grateful user.
> > > > There is a man pager that might be of minor interest, but that can be
> > > > taken over by the kernel team if they want
> >
> > I do not want to maintain the dying package.
> >
Why is it dying? Do you mean the downward trend on popcon? ISTM that
this speaks of the success of the kernel team in reducing the benefit
of a custom kernel...so a decline in relevancy for the majority of
users (plus a cultural shift), but as I see it it's still core
functionality that facilitates freedom.
Of course anyone can orphan a package at any time for any reason ;-)
> > > [...]
> > >
> > > I assume you're referring to kernel-img.conf. I would be happy to add
> > > that to linux-base, with some clarification of which settings are
> > > understood by which packages.
> > >
> > > I take it that you don't want kernel-package to be included in buster,
> > > so can you open an RC bug to say so?
> >
> > Also, could you make sure it won't come back to bullseye or later ?
> >
> > How about Ubuntu? Is it out of your control?
> >
Where are the bugs that make this package unsuitable for release? My
simple use case is not affected by any of
https://bugs.debian.org/cgi-bin/pkgreport.cgi?package=kernel-package
and I don't believe this RC bug is justified.
Is there any reason why any buggy kernel-package features can't be
cut? Removing this package unnecessarily raises the barrier of entry
to building custom kernels...a killer feature of Debian--easy custom
kernels under package management. Of course, if Debian had a kernel
flavour that met my use case then I might not need kernel-package.
The popcon trend appears to be fairly linear and it seems that there
will still be ~2000 users of the package by the time bullseye is
released.
I've attached my homemade convenience/enhancement wrapper script,
because it's possible that it works around some of the alleged unnamed
issues; my experience with this package has been *flawless*. This bug
seems to indicate that it's difficult to maintain, so thank you for
your exceptional work on kernel-package.
Sincerely,
Nicholas
#!/bin/bash
# build-kernel builds a linux kernel with the right version of GCC and
# asks you to install it. It also automatically enables concurrency
# for SMP and multicore systems.
#
# Copyright (C) 2011-2018 Nicholas D Steeves <nstee...@gmail.com>
# Distributed under the terms of the GPLv2 license.
# This script depends on sudo, kernel-package, fakeroot, and pv
linux_source="$PWD"
# linux_name=$(git rev-parse --abbrev-ref HEAD)
# linux_name="a custom name"
# email='yourem...@somedomain.com'
tmpfs='/scratch/tmpfs/'
today=$(date +%Y%m%d)
CONCURRENCY_LEVEL=$(grep -c ^processor /proc/cpuinfo | sed 's/^0$/1/')
CCV=$(cat /proc/version | sed 's/.*gcc version //' | cut -c 1)
MAKEFLAGS="CC=gcc-$CCV"
if [ ! -x "$(command -v gcc-$CCV)" ]; then
echo "A suitable gcc version was not found"
echo "Please install gcc-$CCV before continuing"
exit 1
fi
export CONCURRENCY_LEVEL
export MAKEFLAGS
cd "$linux_source" && make oldconfig || exit $?
# Take this path if a tmpfs has been defined
if [ -n "${tmpfs+x}" ]; then
bytes=$(du -bc ./* | tail -1 | awk '{print $1}')
sudo mkdir -p "$tmpfs"
sudo chown root:sudo "$tmpfs"
sudo chmod 770 "$tmpfs"
sudo mount none -t tmpfs "$tmpfs"
mkdir "$tmpfs"/linux || exit $?
tar cf - ./* .config | pv -pabet -s"$bytes" | (cd $tmpfs/linux && tar xfp
-) \
|| exit $?
cd "$tmpfs"/linux || exit $?
time fakeroot make-kpkg binary-arch \
--initrd --append-to-version=."$today" || exit $?
echo
for i in VERSION PATCHLEVEL SUBLEVEL; do
tmp=$(grep ^$i Makefile | awk '{print $3}')
export "${i}"=$tmp
done
unset tmp
while true; do
printf "Would you like to install the new kernel? (yes/no) "
read choice
case "$choice" in
yes | y | Yes | YES )
break
;;
no | n | No | NO )
echo "Exiting without installing the new kernel"
sync; sleep 3
sudo umount "$tmpfs"
exit 1
;;
* )
printf "Please answer yes or no\n\n"
;;
esac
done
sudo dpkg -i \
"$tmpfs"/linux-headers-${VERSION}.${PATCHLEVEL}.${SUBLEVEL}.${today}_*.deb \
"$tmpfs"/linux-image-${VERSION}.${PATCHLEVEL}.${SUBLEVEL}.${today}_*.deb \
&& (sync; sleep 3; sudo umount "$tmpfs"/)
echo "Reboot to activate new kernel"
# Take the following path if a tmpfs has not been defined
else
time fakeroot make-kpkg binary-arch --initrd \
--append-to-version=."${today}" || exit $?
echo
for i in VERSION PATCHLEVEL SUBLEVEL; do
tmp=$(grep ^$i Makefile | awk '{print $3}')
export "${i}"=$tmp
done
while true; do
printf "Would you like to install the new kernel? (yes/no) "
read choice
case "$choice" in
yes | y | Yes | YES )
break
;;
no | n | No | NO )
echo "Exiting without installing the new kernel"
cd "$linux_source"; sync; sleep 3
sudo umount "$tmpfs"
exit 1
;;
* )
printf "Please answer yes or no\n\n"
;;
esac
done
sudo dpkg -i \
"$linux_source"/../linux-headers-${VERSION}.${PATCHLEVEL}.${SUBLEVEL}.${today}_*.deb
\
"$linux_source"/../linux-image-${VERSION}.${PATCHLEVEL}.${SUBLEVEL}.${today}_*.deb
\
&& (cd "$linux_source"; sync; sleep 3; sudo umount "$tmpfs")
echo "Reboot to activate new kernel"
fi
signature.asc
Description: PGP signature
