You are not logged in.

#1 2017-11-10 08:10:31

andreas_baumann
Administrator
From: Zurich, Switzerland
Registered: 2017-08-10
Posts: 601
Website

core/linux-lts

Cannot use CONFIG_CC_STACKPROTECTOR_STRONG: -fstack-protector-strong not supported by compiler
  SYSHDR  arch/x86/entry/syscalls/../../include/generated/uapi/asm/unistd_x32.h
make: *** [Makefile:1069: prepare-compiler-check] Error 1

Debian says: https://bugs.debian.org/cgi-bin/bugrepo … bug=841368

So I'll try to do that.. :-)

Offline

#2 2017-11-10 08:19:02

andreas_baumann
Administrator
From: Zurich, Switzerland
Registered: 2017-08-10
Posts: 601
Website

Re: core/linux-lts

mmh. So I have to remove the CONFIG_CC_STACKPROTECTOR_STRONG from the kernel options? No way!
And why is this happening suddenly? Recent gcc update?

Offline

#3 2017-11-10 08:25:16

andreas_baumann
Administrator
From: Zurich, Switzerland
Registered: 2017-08-10
Posts: 601
Website

Re: core/linux-lts

Ah. The kernel config was shared before between 64 and 32 bit (same config for both architectures). In packages
commit 0d035f02333036f92a6bcf3825a77619145d55aa I see no longer the i686 part. So, I suggest we keep
our own kernel configs from now on.

Offline

#4 2017-11-10 08:31:23

andreas_baumann
Administrator
From: Zurich, Switzerland
Registered: 2017-08-10
Posts: 601
Website

Re: core/linux-lts

This brings up another issue: when the 64-bit-guys (no offence meant) throw out more 32-bit stuff in PKGBUILD, I expect to see tons of broken packages here.
So basically every package has to be checked for differences which could be 32-bit specific.

Offline

#5 2017-11-10 08:35:27

andreas_baumann
Administrator
From: Zurich, Switzerland
Registered: 2017-08-10
Posts: 601
Website

Re: core/linux-lts

For instance:

diff ../work/repos/packages/linux-lts/trunk/PKGBUILD ../work/repos/packages/linux-lts/repos/core-i686/PKGBUILD

7c7
< pkgver=4.9.61
---
> pkgver=4.9.60
9c9
< arch=('x86_64')
---
> arch=('i686' 'x86_64')
17c17
<         'config'
---
>         'config' 'config.x86_64'
26c26
<             '73abd713dfbc954cc3d2eebfd35370629b3e6e09e2a539dba2bd608f03b5947b'
---
>             '10f5cce03294a7bec09ffdfae3946d0f3e6cd7aced9044aa35c3f88d2cb62f6a'
27a28
>             'f5611ebd9d9ff0aad420cb74857cd6b034ebc1d3fea480bca8a43610780907d5'
52c53,57
<   cat "${srcdir}/config" > ./.config
---
>   if [ "${CARCH}" = "x86_64" ]; then
>     cat "${srcdir}/config.x86_64" > ./.config
>   else
>     cat "${srcdir}/config" > ./.config
>   fi
183a189,192
>
>   if [ "${CARCH}" = "i686" ]; then
>     cp arch/${KARCH}/Makefile_32.cpu "${pkgdir}/usr/lib/modules/${_kernver}/build/arch/${KARCH}/"
>   fi

This means all those snippets have to be added to the vanilla PKGBUILD (trunk) in our diff PKGBUILD's.

Offline

#6 2017-11-17 15:50:11

deep42thought
Administrator
From: Jena, Germany
Registered: 2017-06-17
Posts: 378

Re: core/linux-lts

Yes, but I would act, as soon as things start breaking.
We can always look in the git history, what the difference between x86_64 and i686 was when both were still present (usually it's one PKGBUILD with a few switches) and then put all the differences in our modification-repository.

Offline

#7 2017-11-17 15:52:10

andreas_baumann
Administrator
From: Zurich, Switzerland
Registered: 2017-08-10
Posts: 601
Website

Re: core/linux-lts

The snippets above are in the diff PKGBUILD, the config-i686 is a copy from the upstream repo.

Offline

Board footer

Powered by FluxBB