[opensuse-factory] KMP build/packaging failure for i586
Hi, I encounter a strange error building 3rd party KMP rpms. If you look at the x86_64 build : https://build.opensuse.org/package/live_build_log/openSUSE:Factory/openafs/s... everything if fine. Using the same spec the 32bit i586 build fails in the actual packaging step: https://build.opensuse.org/package/live_build_log/openSUSE:Factory/openafs/s... ... """ [ 2373s] Processing files: openafs-kmp-default-1.8.0_krpmbuild-1.19.i586 [ 2373s] error: File must begin with "/": * [ 2373s] error: File must begin with "/": Restart [ 2373s] error: File must begin with "/": config... [ 2373s] error: File must begin with "/": * [ 2373s] error: File must begin with "/": * [ 2373s] error: File must begin with "/": * [ 2373s] error: File must begin with "/": Linux/x86 [ 2373s] error: File must begin with "/": 4.17.1 [ 2373s] error: File must begin with "/": Kernel [ 2373s] error: File must begin with "/": Configuration [ 2373s] error: File must begin with "/": * [ 2373s] error: File must begin with "/": 64-bit [ 2373s] error: File must begin with "/": kernel [ 2373s] error: File must begin with "/": (64BIT) [ 2373s] error: File must begin with "/": [Y/n/?] [ 2373s] error: File must begin with "/": (NEW) [ 2373s] error: File must begin with "/": * """... Apparently, for the i586-case the automatically created file-list for the KMP is overwritten by the output of the kernel-configuration. Any ideas how to debug this ? Many thanks, Christof -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Mon, 25 Jun 2018 17:11:32 +0200 Christof Hanke <christof.hanke@mpcdf.mpg.de> wrote:
Hi,
Hello, this is bug 1098050 Please update your package to work with the new kernels. You can see the required changes in the SRs references in the bugzilla. Thanks Michal -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Mon, 25 Jun 2018 17:17:43 +0200, Michal Suchánek wrote:
On Mon, 25 Jun 2018 17:11:32 +0200 Christof Hanke <christof.hanke@mpcdf.mpg.de> wrote:
Hi,
Hello,
this is bug 1098050
Please update your package to work with the new kernels.
You can see the required changes in the SRs references in the bugzilla.
Hm, so this hits every KMP build? That's too bad... Can we somehow trick it away, e.g. by giving ARCH=i586 in /usr/src/linux-obj/i586/*/Makefile? Takashi -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Mon, 25 Jun 2018 17:46:42 +0200 Takashi Iwai <tiwai@suse.de> wrote:
On Mon, 25 Jun 2018 17:17:43 +0200, Michal Suchánek wrote:
On Mon, 25 Jun 2018 17:11:32 +0200 Christof Hanke <christof.hanke@mpcdf.mpg.de> wrote:
Hi,
Hello,
this is bug 1098050
Please update your package to work with the new kernels.
You can see the required changes in the SRs references in the bugzilla.
Hm, so this hits every KMP build? That's too bad...
Every KMP that builds on i586, specifically. Other archs happen to works.
Can we somehow trick it away, e.g. by giving ARCH=i586 in /usr/src/linux-obj/i586/*/Makefile?
Probably. However, this is an upstream requirement and is not going to go away in foreseeable future so we might as well fix the KMPs. Thanks Michal -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Mon, 25 Jun 2018 17:57:42 +0200, Michal Suchánek wrote:
On Mon, 25 Jun 2018 17:46:42 +0200 Takashi Iwai <tiwai@suse.de> wrote:
On Mon, 25 Jun 2018 17:17:43 +0200, Michal Suchánek wrote:
On Mon, 25 Jun 2018 17:11:32 +0200 Christof Hanke <christof.hanke@mpcdf.mpg.de> wrote:
Hi,
Hello,
this is bug 1098050
Please update your package to work with the new kernels.
You can see the required changes in the SRs references in the bugzilla.
Hm, so this hits every KMP build? That's too bad...
Every KMP that builds on i586, specifically. Other archs happen to works.
So it hits most of KMPs :)
Can we somehow trick it away, e.g. by giving ARCH=i586 in /usr/src/linux-obj/i586/*/Makefile?
Probably. However, this is an upstream requirement and is not going to go away in foreseeable future so we might as well fix the KMPs.
Yes but adding ARCH= would be exactly the fix for all KMPs. What else needed in each KMP side? And I'm not sure whether this is the intended behavior. This kind of thing may be treated as a regression, too. BTW, is the affecting upstream commit 104daea149c4 kconfig: reference environment variables directly and remove 'option env=' ? thanks, Takashi -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Mon, 25 Jun 2018 18:11:43 +0200 Takashi Iwai <tiwai@suse.de> wrote:
On Mon, 25 Jun 2018 17:57:42 +0200, Michal Suchánek wrote:
On Mon, 25 Jun 2018 17:46:42 +0200 Takashi Iwai <tiwai@suse.de> wrote:
On Mon, 25 Jun 2018 17:17:43 +0200, Michal Suchánek wrote:
On Mon, 25 Jun 2018 17:11:32 +0200 Christof Hanke <christof.hanke@mpcdf.mpg.de> wrote:
Hi,
Hello,
this is bug 1098050
Please update your package to work with the new kernels.
You can see the required changes in the SRs references in the bugzilla.
Hm, so this hits every KMP build? That's too bad...
Every KMP that builds on i586, specifically. Other archs happen to works.
So it hits most of KMPs :)
Can we somehow trick it away, e.g. by giving ARCH=i586 in /usr/src/linux-obj/i586/*/Makefile?
Probably. However, this is an upstream requirement and is not going to go away in foreseeable future so we might as well fix the KMPs.
Yes but adding ARCH= would be exactly the fix for all KMPs. What else needed in each KMP side?
And I'm not sure whether this is the intended behavior. This kind of thing may be treated as a regression, too.
BTW, is the affecting upstream commit 104daea149c4 kconfig: reference environment variables directly and remove 'option env=' ?
I suppose it might be a bug. We have # Select 32 or 64 bit config 64BIT bool "64-bit kernel" if ARCH = "x86" default ARCH != "i386" ---help--- Say yes to build a 64-bit kernel - formerly known as x86_64 Say no to build a 32-bit kernel - formerly known as i386 So the option has a description and the value should be saved even after f467c5640c29ad258c3cd8186a776c82fc3b8057 ("kconfig: only write '# CONFIG_FOO is not set' for visible symbols") but it is not. Thanks Michal -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tue, 26 Jun 2018 10:13:35 +0200, Michal Suchánek wrote:
On Mon, 25 Jun 2018 18:11:43 +0200 Takashi Iwai <tiwai@suse.de> wrote:
On Mon, 25 Jun 2018 17:57:42 +0200, Michal Suchánek wrote:
On Mon, 25 Jun 2018 17:46:42 +0200 Takashi Iwai <tiwai@suse.de> wrote:
On Mon, 25 Jun 2018 17:17:43 +0200, Michal Suchánek wrote:
On Mon, 25 Jun 2018 17:11:32 +0200 Christof Hanke <christof.hanke@mpcdf.mpg.de> wrote:
Hi,
Hello,
this is bug 1098050
Please update your package to work with the new kernels.
You can see the required changes in the SRs references in the bugzilla.
Hm, so this hits every KMP build? That's too bad...
Every KMP that builds on i586, specifically. Other archs happen to works.
So it hits most of KMPs :)
Can we somehow trick it away, e.g. by giving ARCH=i586 in /usr/src/linux-obj/i586/*/Makefile?
Probably. However, this is an upstream requirement and is not going to go away in foreseeable future so we might as well fix the KMPs.
Yes but adding ARCH= would be exactly the fix for all KMPs. What else needed in each KMP side?
And I'm not sure whether this is the intended behavior. This kind of thing may be treated as a regression, too.
BTW, is the affecting upstream commit 104daea149c4 kconfig: reference environment variables directly and remove 'option env=' ?
I suppose it might be a bug. We have
# Select 32 or 64 bit config 64BIT bool "64-bit kernel" if ARCH = "x86" default ARCH != "i386" ---help--- Say yes to build a 64-bit kernel - formerly known as x86_64 Say no to build a 32-bit kernel - formerly known as i386
So the option has a description and the value should be saved even after f467c5640c29ad258c3cd8186a776c82fc3b8057 ("kconfig: only write '# CONFIG_FOO is not set' for visible symbols") but it is not.
Ah that commit, thanks, now it's clearer. So yes I agree that it's the unexpected behavior. Care to report it to upstream? thanks, Takashi -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tuesday, 26 June 2018 10:13 Michal Suchánek wrote:
I suppose it might be a bug. We have
# Select 32 or 64 bit config 64BIT bool "64-bit kernel" if ARCH = "x86" default ARCH != "i386" ---help--- Say yes to build a 64-bit kernel - formerly known as x86_64 Say no to build a 32-bit kernel - formerly known as i386
So the option has a description and the value should be saved even after f467c5640c29ad258c3cd8186a776c82fc3b8057 ("kconfig: only write '# CONFIG_FOO is not set' for visible symbols") but it is not.
The trick seems to be the `if ARCH = "x86"' above. There are three possibilities: - with ARCH not set, it defaults to "x86" and "make *config" asks for the value (and writes the result) - with ARCH=x86_64, "make *config" does not ask, defaults to "y" and writes the "result" - with ARCH=i386, "make *config" does not ask, defaults to "n" and does not write the result (as the symbol is not "visible") However, we are going to have much bigger problem starting with 4.18-rc1 as there even running "make oldconfig" with ARCH set may result in different .config than running it on the actual architecture. Michal -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tue, 26 Jun 2018 10:26:22 +0200, Michal Kubecek wrote:
On Tuesday, 26 June 2018 10:13 Michal Suchánek wrote:
I suppose it might be a bug. We have
# Select 32 or 64 bit config 64BIT bool "64-bit kernel" if ARCH = "x86" default ARCH != "i386" ---help--- Say yes to build a 64-bit kernel - formerly known as x86_64 Say no to build a 32-bit kernel - formerly known as i386
So the option has a description and the value should be saved even after f467c5640c29ad258c3cd8186a776c82fc3b8057 ("kconfig: only write '# CONFIG_FOO is not set' for visible symbols") but it is not.
The trick seems to be the `if ARCH = "x86"' above. There are three possibilities:
- with ARCH not set, it defaults to "x86" and "make *config" asks for the value (and writes the result) - with ARCH=x86_64, "make *config" does not ask, defaults to "y" and writes the "result" - with ARCH=i386, "make *config" does not ask, defaults to "n" and does not write the result (as the symbol is not "visible")
Argh, right, thanks for pointing out. So assuming that passing ARCH= is now mandatory, I believe we should handle it in Makefile, not by passing from each KMP.
However, we are going to have much bigger problem starting with 4.18-rc1 as there even running "make oldconfig" with ARCH set may result in different .config than running it on the actual architecture.
Oh that's messy... How can it happen? Takashi -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tuesday, 26 June 2018 10:53 Takashi Iwai wrote:
On Tue, 26 Jun 2018 10:26:22 +0200, Michal Kubecek wrote:
However, we are going to have much bigger problem starting with 4.18-rc1 as there even running "make oldconfig" with ARCH set may result in different .config than running it on the actual architecture.
Oh that's messy... How can it happen?
There is a more detailed e-mail about this in SUSE kernel list (subject "Kconfig changes in v4.18-rc1"). The short story is that Kconfig now writes also info about gcc version and capabilities into .config so that using different gcc when updating configs than one used later for build may be a problem. And I also found than updating ppc64le/default config (master branch) with gcc 8.1.1 on x86_64 (with ARCH=powerpc) - which is what run_oldconfig.sh does - results in different .config than when I do the same on ppc64le machine. I'm going to do some research if we can work around this without having to use at least a crosscompiler. Michal -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Jun 26 2018, Michal Kubecek <mkubecek@suse.cz> wrote:
However, we are going to have much bigger problem starting with 4.18-rc1 as there even running "make oldconfig" with ARCH set may result in different .config than running it on the actual architecture.
If you create .config with ARCH=x86 then running oldconfig with ARCH unset works as expected. Andreas. -- Andreas Schwab, SUSE Labs, schwab@suse.de GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE 1748 E4D4 88E3 0EEA B9D7 "And now for something completely different." -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tuesday, 26 June 2018 11:08 Andreas Schwab wrote:
On Jun 26 2018, Michal Kubecek <mkubecek@suse.cz> wrote:
However, we are going to have much bigger problem starting with 4.18-rc1 as there even running "make oldconfig" with ARCH set may result in different .config than running it on the actual architecture. If you create .config with ARCH=x86 then running oldconfig with ARCH unset works as expected.
In this particular case. Not in general. Michal Kubecek -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Jun 26 2018, Michal Kubecek <mkubecek@suse.cz> wrote:
On Tuesday, 26 June 2018 11:08 Andreas Schwab wrote:
On Jun 26 2018, Michal Kubecek <mkubecek@suse.cz> wrote:
However, we are going to have much bigger problem starting with 4.18-rc1 as there even running "make oldconfig" with ARCH set may result in different .config than running it on the actual architecture. If you create .config with ARCH=x86 then running oldconfig with ARCH unset works as expected.
In this particular case. Not in general.
How general? Andreas. -- Andreas Schwab, SUSE Labs, schwab@suse.de GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE 1748 E4D4 88E3 0EEA B9D7 "And now for something completely different." -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tuesday, 26 June 2018 11:14 Andreas Schwab wrote:
On Jun 26 2018, Michal Kubecek <mkubecek@suse.cz> wrote:
On Tuesday, 26 June 2018 11:08 Andreas Schwab wrote:
On Jun 26 2018, Michal Kubecek <mkubecek@suse.cz> wrote:
However, we are going to have much bigger problem starting with 4.18-rc1 as there even running "make oldconfig" with ARCH set may result in different .config than running it on the actual architecture.
If you create .config with ARCH=x86 then running oldconfig with ARCH unset works as expected.
In this particular case. Not in general.
How general?
See my earlier e-mail. Check out the "master" branch, run "ARCH=powerpc make oldconfig" with ppc64le/default config on x86_64 and on ppc64le. The results are different. Michal Kubecek -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Jun 26 2018, Michal Kubecek <mkubecek@suse.cz> wrote:
See my earlier e-mail. Check out the "master" branch, run "ARCH=powerpc make oldconfig" with ppc64le/default config on x86_64 and on ppc64le. The results are different.
Are the differences all dependencies of GCC_VERSION? That's a real problem, I agree. Andreas. -- Andreas Schwab, SUSE Labs, schwab@suse.de GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE 1748 E4D4 88E3 0EEA B9D7 "And now for something completely different." -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tue, Jun 26, 2018 at 11:53:57AM +0200, Andreas Schwab wrote:
On Jun 26 2018, Michal Kubecek <mkubecek@suse.cz> wrote:
See my earlier e-mail. Check out the "master" branch, run "ARCH=powerpc make oldconfig" with ppc64le/default config on x86_64 and on ppc64le. The results are different.
Are the differences all dependencies of GCC_VERSION? That's a real problem, I agree.
Not only version, there seems to be a problem with capability detection. This is the diff: diff --git a/config/ppc64le/default b/config/ppc64le/default index 3d2b2a11fd47..ee72f4d3c5c8 100644 --- a/config/ppc64le/default +++ b/config/ppc64le/default @@ -4,7 +4,7 @@ # # -# Compiler: gcc (SUSE Linux) 8.1.1 20180614 [gcc-8-branch revision 261584] +# Compiler: gcc (SUSE Linux) 8.1.1 20180613 [gcc-8-branch revision 261538] # CONFIG_PPC64=y @@ -275,7 +275,6 @@ CONFIG_KPROBES=y CONFIG_JUMP_LABEL=y # CONFIG_STATIC_KEYS_SELFTEST is not set CONFIG_OPTPROBES=y -CONFIG_KPROBES_ON_FTRACE=y CONFIG_UPROBES=y CONFIG_HAVE_64BIT_ALIGNED_ACCESS=y CONFIG_ARCH_USE_BUILTIN_BSWAP=y @@ -544,7 +543,6 @@ CONFIG_COREDUMP=y CONFIG_HUGETLB_PAGE_SIZE_VARIABLE=y CONFIG_PPC_TRANSACTIONAL_MEM=y # CONFIG_LD_HEAD_STUB_CATCH is not set -CONFIG_MPROFILE_KERNEL=y CONFIG_HOTPLUG_CPU=y CONFIG_ARCH_CPU_PROBE_RELEASE=y CONFIG_ARCH_ENABLE_MEMORY_HOTPLUG=y @@ -6610,7 +6608,6 @@ CONFIG_NOP_TRACER=y CONFIG_HAVE_FUNCTION_TRACER=y CONFIG_HAVE_FUNCTION_GRAPH_TRACER=y CONFIG_HAVE_DYNAMIC_FTRACE=y -CONFIG_HAVE_DYNAMIC_FTRACE_WITH_REGS=y CONFIG_HAVE_FTRACE_MCOUNT_RECORD=y CONFIG_HAVE_SYSCALL_TRACEPOINTS=y CONFIG_TRACER_MAX_TRACE=y @@ -6642,7 +6639,6 @@ CONFIG_UPROBE_EVENTS=y CONFIG_BPF_EVENTS=y CONFIG_PROBE_EVENTS=y CONFIG_DYNAMIC_FTRACE=y -CONFIG_DYNAMIC_FTRACE_WITH_REGS=y CONFIG_FUNCTION_PROFILER=y CONFIG_FTRACE_MCOUNT_RECORD=y # CONFIG_FTRACE_STARTUP_TEST is not set @@ -7009,8 +7005,6 @@ CONFIG_VHOST_SCSI=m CONFIG_VHOST_VSOCK=m CONFIG_VHOST=m # CONFIG_VHOST_CROSS_ENDIAN_LEGACY is not set -CONFIG_HAVE_LIVEPATCH=y -CONFIG_LIVEPATCH=y CONFIG_SUSE_KERNEL=y # And now it seems there may be also difference for aarch64. :-( Michal Kubecek -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Jun 26 2018, Michal Kubecek <mkubecek@suse.cz> wrote:
@@ -544,7 +543,6 @@ CONFIG_COREDUMP=y CONFIG_HUGETLB_PAGE_SIZE_VARIABLE=y CONFIG_PPC_TRANSACTIONAL_MEM=y # CONFIG_LD_HEAD_STUB_CATCH is not set -CONFIG_MPROFILE_KERNEL=y
That's another config that depends on a proper cross compiler (see arch/powerpc/tools/gcc-check-mprofile-kernel.sh). Andreas. -- Andreas Schwab, SUSE Labs, schwab@suse.de GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE 1748 E4D4 88E3 0EEA B9D7 "And now for something completely different." -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tue, 26 Jun 2018 11:14:49 +0200 Andreas Schwab <schwab@suse.de> wrote:
On Jun 26 2018, Michal Kubecek <mkubecek@suse.cz> wrote:
On Tuesday, 26 June 2018 11:08 Andreas Schwab wrote:
On Jun 26 2018, Michal Kubecek <mkubecek@suse.cz> wrote:
However, we are going to have much bigger problem starting with 4.18-rc1 as there even running "make oldconfig" with ARCH set may result in different .config than running it on the actual architecture. If you create .config with ARCH=x86 then running oldconfig with ARCH unset works as expected.
In this particular case. Not in general.
How general?
The kernel configuration and build depends on environment. Previously the relevant environment variables were saved in config so the config was self-contained. Now you have a link to source and object directory from /lib/modules/$(uname -r) but the link does not have the environment anywhere anymore so in general you cannot build modules with what you have. Similarly you need the environment to interpret /proc/config.gz but it is not available anywhere. Thanks Michal -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Jun 26 2018, Michal Suchánek <msuchanek@suse.de> wrote:
I suppose it might be a bug. We have
# Select 32 or 64 bit config 64BIT bool "64-bit kernel" if ARCH = "x86" default ARCH != "i386" ---help--- Say yes to build a 64-bit kernel - formerly known as x86_64 Say no to build a 32-bit kernel - formerly known as i386
So the option has a description and the value should be saved even after f467c5640c29ad258c3cd8186a776c82fc3b8057 ("kconfig: only write '# CONFIG_FOO is not set' for visible symbols") but it is not.
Only if ARCH=x86, but the .config was created with ARCH=i386. I think ARCH should be set to x86 for both x86_64 and i?86. Andreas. -- Andreas Schwab, SUSE Labs, schwab@suse.de GPG Key fingerprint = 0196 BAD8 1CE9 1970 F4BE 1748 E4D4 88E3 0EEA B9D7 "And now for something completely different." -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
Hi, On Mon, 25 Jun 2018 17:17:43 +0200 Michal Suchánek <msuchanek@suse.de> wrote:
On Mon, 25 Jun 2018 17:11:32 +0200 Christof Hanke <christof.hanke@mpcdf.mpg.de> wrote:
this is bug 1098050
Please update your package to work with the new kernels.
You can see the required changes in the SRs references in the bugzilla.
Unfortunately, the actual make command is in a script and thus I cannot use the macro. However, passing ARCH=i386 to the make-command is apparently not enough: I see in the build.log (from inside the chroot): abuild# grep \ modules /.build.log | grep make [ 503s] make -C /usr/src/linux-obj/i586/default ARCH=i386 M=/home/abuild/rpmbuild/BUILD/openafs-1.6.22.3/obj/default/src/libafs/MODLOAD-4.17.2-1-default-MP modules 2>&1 \ [ 571s] + make -C /usr/src/linux-obj/i586/default ARCH=i386 modules_install M=/home/abuild/rpmbuild/BUILD/openafs-1.6.22.3/obj/default/src/libafs/MODLOAD-4.17.2-1-default-MP But still it does not work. Also on OBS it's not working. See eg. https://build.opensuse.org/package/live_build_log/home:hauky:branches:filesy... Besides, I noticed that the macro linux_make_arch is not set (I echo it in the beginning and it is empty). Any ideas ? Thanks, Christof
Thanks
Michal
-- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tue, 26 Jun 2018 12:07:41 +0200 Christof Hanke <christof.hanke@mpcdf.mpg.de> wrote:
Hi,
On Mon, 25 Jun 2018 17:17:43 +0200 Michal Suchánek <msuchanek@suse.de> wrote:
On Mon, 25 Jun 2018 17:11:32 +0200 Christof Hanke <christof.hanke@mpcdf.mpg.de> wrote:
this is bug 1098050
Please update your package to work with the new kernels.
You can see the required changes in the SRs references in the bugzilla.
Unfortunately, the actual make command is in a script and thus I cannot use the macro.
However, passing ARCH=i386 to the make-command is apparently not enough:
I see in the build.log (from inside the chroot):
abuild# grep \ modules /.build.log | grep make [ 503s] make -C /usr/src/linux-obj/i586/default ARCH=i386 M=/home/abuild/rpmbuild/BUILD/openafs-1.6.22.3/obj/default/src/libafs/MODLOAD-4.17.2-1-default-MP modules 2>&1 \ [ 571s] + make -C /usr/src/linux-obj/i586/default ARCH=i386 modules_install M=/home/abuild/rpmbuild/BUILD/openafs-1.6.22.3/obj/default/src/libafs/MODLOAD-4.17.2-1-default-MP
But still it does not work. Also on OBS it's not working.
See eg. https://build.opensuse.org/package/live_build_log/home:hauky:branches:filesy...
Besides, I noticed that the macro linux_make_arch is not set (I echo it in the beginning and it is empty).
Any ideas ?
You need the kernel from KErnel:stable. The macro is not defined in Factory yet. Thanks Michal -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On Tue, 26 Jun 2018 14:40:31 +0200 Michal Suchánek <msuchanek@suse.de> wrote:
On Tue, 26 Jun 2018 12:07:41 +0200 Christof Hanke <christof.hanke@mpcdf.mpg.de> wrote:
[...] [...] [...] [...] [...]
You need the kernel from KErnel:stable. The macro is not defined in Factory yet.
Thanks, now I got it working. Christof -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
On 06/25/2018 10:11 AM, Christof Hanke wrote:
Hi,
I encounter a strange error building 3rd party KMP rpms. If you look at the x86_64 build :
https://build.opensuse.org/package/live_build_log/openSUSE:Factory/openafs/s...
everything if fine. Using the same spec the 32bit i586 build fails in the actual packaging step:
https://build.opensuse.org/package/live_build_log/openSUSE:Factory/openafs/s...
... """ [ 2373s] Processing files: openafs-kmp-default-1.8.0_krpmbuild-1.19.i586 [ 2373s] error: File must begin with "/": * [ 2373s] error: File must begin with "/": Restart [ 2373s] error: File must begin with "/": config... [ 2373s] error: File must begin with "/": * [ 2373s] error: File must begin with "/": * [ 2373s] error: File must begin with "/": * [ 2373s] error: File must begin with "/": Linux/x86 [ 2373s] error: File must begin with "/": 4.17.1 [ 2373s] error: File must begin with "/": Kernel [ 2373s] error: File must begin with "/": Configuration [ 2373s] error: File must begin with "/": * [ 2373s] error: File must begin with "/": 64-bit [ 2373s] error: File must begin with "/": kernel [ 2373s] error: File must begin with "/": (64BIT) [ 2373s] error: File must begin with "/": [Y/n/?] [ 2373s] error: File must begin with "/": (NEW) [ 2373s] error: File must begin with "/": * """...
Apparently, for the i586-case the automatically created file-list for the KMP is overwritten by the output of the kernel-configuration.
Any ideas how to debug this ?
For VirtualBox, the fix was to add the %{?linux_make_arch} macro to the make line. Thus, it became make -j2 -C %{_prefix}/src/linux-obj/%{_target_cpu}/$flavor %{?linux_make_arch}\ modules M=$PWD/modules_build_dir/$flavor/$module_name Larry -- To unsubscribe, e-mail: opensuse-factory+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-factory+owner@opensuse.org
participants (6)
-
Andreas Schwab -
Christof Hanke -
Larry Finger -
Michal Kubecek -
Michal Suchánek -
Takashi Iwai