Arch Linux User Repository

Thanks @glitsj16

I applied it and rebased. In doing so I found that I had to change 0004-hrtimer-Replace-all-schedule-timeout-1-with-schedule.patch ... can you take a look at that and let me know if I did it correctly? Building 6.2.8 was successful.

https://github.com/graysky2/linux-patches/commit/a27ac12de07749d0c935249d5031514b6cbd1a35

I am getting a similar error. I need some help rebasing 6.2 into the patches.

@graysky Here's my attempt at fixing the rebase: https://gist.github.com/glitsj16/46ba4eebbde9727577c2a2075dfda3c1 - tested with 6.2.7 (6.2.8 is building right now).

Unfortunately my skill levels are far too low for that and my time currently is too limited to acquire those skills. I can edit PKGBUILD's just fine, but everything else is way above me.

I am getting a similar error. I need some help rebasing 6.2 into the patches.

  CC [M]  arch/x86/kvm/x86.o
  CC      crypto/aes_generic.o
kernel/sysctl.c:1666:35: error: ‘sysctl_numa_balancing’ undeclared here (not in a function); did you mean ‘sysctl_numa_balancing_mode’?
 1666 |                 .proc_handler   = sysctl_numa_balancing,
      |                                   ^~~~~~~~~~~~~~~~~~~~~
      |                                   sysctl_numa_balancing_mode
  CC      kernel/time/tick-sched.o
  CC [M]  arch/x86/kvm/emulate.o
  CC      lib/zstd/zstd_common_module.o
kernel/sysctl.c:1672:36: error: ‘sysctl_numa_balancing_promote_rate_limit’ undeclared here (not in a function)
 1672 |                 .data           = &sysctl_numa_balancing_promote_rate_limit,
      |                                    ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
  CC      drivers/pinctrl/intel/pinctrl-emmitsburg.o
make[2]: *** [scripts/Makefile.build:252: kernel/sysctl.o] Error 1

I am getting the following error when trying to build the update to 6.2 (coming from 6.1.12). I already tried a clean build as well as manually removing the build folder (rm -rf ~/.cache/yay/linux-ck/) Never had any issues with this package, using it since 4.something on different machines.

  CC [M]  drivers/iio/adc/sd_adc_modulator.o
  CC [M]  drivers/iio/light/tsl4531.o
  CC [M]  drivers/iio/light/us5182d.o
  CC [M]  drivers/iio/light/vcnl4000.o
  CC [M]  drivers/iio/light/vcnl4035.o
  CC [M]  drivers/iio/light/veml6030.o
  CC [M]  drivers/iio/light/veml6070.o
  CC [M]  drivers/iio/light/vl6180.o
  CC [M]  drivers/iio/light/zopt2201.o
  LD [M]  drivers/iio/adc/xilinx-xadc.o
  LD [M]  drivers/infiniband/hw/hfi1/hfi1.o
make: *** [Makefile:2021: .] Error 2
==> ERROR: A failure occurred in build().
    Aborting...
 -> error making: linux-ck (linux-ck linux-ck-headers)

I am using the 39 (x86_64v4) build option. Any ideas on what is going wrong and how to fix it?

Edit: The same error happens when using the default 36 option.

I think bison should be specified as make dependency:

Hunk #1 succeeded at 95 (offset 3 lines).
Hunk #2 succeeded at 1639 (offset 3 lines).
patching file kernel/time/clockevents.c
patching file kernel/time/hrtimer.c
  YACC    scripts/kconfig/parser.tab.[ch]
/bin/sh: line 1: bison: command not found
make[1]: *** [scripts/Makefile.host:17: scripts/kconfig/parser.tab.h] Error 127
make: *** [Makefile:697: olddefconfig] Error 2
==> ERROR: A failure occurred in prepare().
    Aborting...

I've been tinkering with some custom patches for this package, but what would be the best way to automate the build process, in particular; How do I prevent the half a dozen prompts for drivers, like (GPIO_BT8XX)?

I'm a bit of a novice when it comes to this, so some help would be appreciated!

I appreciate your feedback, and I'll look into the snippet you shared so I can integrate it into my build system. Thank you!! 😊

What you're proposing is too non-standard for most other users IMO. If you can tweak your build script that would be preferable for your edge use case.

For my build script, the strategy is to first make the set of PKGBUILD files needed for the 4 subarches then copy them into a working dir to build. You might try something like that.

makework() {
  TODO="$des/TODO"
  if [[ ! -d "$TODO" ]]; then
    mkdir -p "$TODO" || exit 1
  fi

  cp "$pkg_source/PKGBUILD" "$TODO" || exit 1

  for i in "${buildem[@]}"; do
    if [[ "$i" = "generic" ]]; then
      # generic setup where 36 is the default
      num=36
      # this hit both instances of 'make oldconfig' for they are in an if statement so doesnt matter
      sed -e s/"#groups=('ck-generic')"/"groups=('ck-generic')"/ \
        -e "s/_subarch=/_subarch=36/" <"$TODO/PKGBUILD" >"$TODO/PKGBUILD.$i"
        #-e "s/make oldconfig/yes $num | make oldconfig/g" <"$TODO/PKGBUILD" >"$TODO/PKGBUILD.$i"
            else
              # not generic so everything else
              # num refers to the number added by my gcc patch assuming x86_64's list
              case $i in
                generic-v2)
                  num=37
                  ;;
                generic-v3)
                  num=38
                  ;;
                generic-v4)
                  num=39
                  ;;
                *) exit 0
                  ;;
              esac

              sed -e s"/#groups=('ck-generic')"/"groups=('ck-generic')"/ \
                -e "/^pkgbase=/ s,linux-ck,linux-ck-$i," \
                -e "/depends=('linux-ck'/ s,ck,ck-$i," \
                -e "s/groups=('ck-generic')"/"groups=('ck-$i')"/ \
                -e "s/_subarch=/_subarch=$num/" \
                <"$TODO/PKGBUILD" >"$TODO/PKGBUILD.$i"
    fi
  done

  # only work with PKGBUILD.xxx
  rm -f "$TODO/PKGBUILD"
}

Editing the file makes it harder for automated builds, and it's trivial to add those entries as environment variables on the PKGBUILD. I understand you are not obligated to do so; you have it this way, and anyone who wants it differently can fork it, just like with any other FOSS project.

To make it easy for automated builds, would you be willing to use environment variables on the PKGBUILD to choose the sub-architecture and the compiler?

Thank you for creating the PKGBUILD for this kernel 🤝