Summary: | mate-base/mate-applets: fails to build with cpupower 4.7 | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Toralf Förster <toralf> |
Component: | Current packages | Assignee: | Gentoo MATE Desktop <mate> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | atoth, floppym, joakim.tjernlund, phmagic, russell |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
See Also: | https://bugzilla.kernel.org/show_bug.cgi?id=153161 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
emerge-history.txt
environment mate-base:mate-applets-1.12.1-r2:20160910-234337.log |
Description
Toralf Förster
2016-09-11 08:11:45 UTC
Created attachment 445434 [details]
emerge-history.txt
Created attachment 445436 [details]
environment
Created attachment 445438 [details]
mate-base:mate-applets-1.12.1-r2:20160910-234337.log
I can confirm but this happens only with 4.7 linux-headers, since cpuidle.h is not present anymore As usual, thanks for the report Toralf. I've reverted the addition of this new patch until we can do more testing and work with upstream to get a proper solution. The current bug was caused by a bad patch intended to resolve this issue. After the reversion of the aforementioned patch, and some research into the issue, we appear to be waiting on an update from cpupower's maintainers. *** Bug 593654 has been marked as a duplicate of this bug. *** 1.14 and 9999 have a temporary revert of the upstream fix until we see a resolution of the bug at kernel.org. For the time being, we are requiring <cpupower-4.7. cpupower 4.9 has been released a while ago - does that include a fix? (In reply to Mihai Moldovan from comment #9) > cpupower 4.9 has been released a while ago - does that include a fix? Seems to be solved since cpupower-4.9 puts into public the function cpupower_is_cpu_online But you must disable the patch in mate-applets ebuild: # PATCHES=( "${FILESDIR}/${PN}-1.14.1-revert-upstream-cpupower-4.7-fix.patch" ) This compiles then. I can confirm that disabling the patch works fine (for ~amd64), provided that the line <sys-power/cpupower-4.7 is changed to sys-power/cpupower:0 as well (and that cpupower is upgraded before trying to recompile mate-applets). Isn't this good enough for releasing a mate-applets-1.18.1-r1 with these changes implemented? It would be nice not to have to edit the ebuild after every sync... (In reply to pste from comment #11) > I can confirm that disabling the patch works fine (for ~amd64), provided > that the line > > <sys-power/cpupower-4.7 > > is changed to > > sys-power/cpupower:0 > > as well (and that cpupower is upgraded before trying to recompile > mate-applets). I also confirm, that latest mate-applets in tree works fine after disabling the patch with cpupower-4.9. Simply removing the patch is not a solution, and the core issue is not fixed. Firstly, if you emerge cpupower-4.9, and then mate-applets-1.18.1, it still breaks, if you have stable linux-headers, which are 4.4, so based on the upstream based code, it uses linux's version to determine whether to use cpupower or cpufreq, and that breaks. Secondly, even if you force people to upgrade to unstable linux-headers (>=4.7) to force it to chose the right code segment, we still end up with: * QA Notice: Package triggers severe warnings which indicate that it * may exhibit random runtime failures. * cpufreq-monitor-libcpufreq.c:117:7: warning: implicit declaration of function ‘cpupower_is_cpu_online’ [-Wimplicit-function-declaration] because there is no public header that includes cpupower_is_cpu_online. See the upstream kernel bug, which still isn't resolved. (In reply to NP-Hardass from comment #13) > Simply removing the patch is not a solution, and the core issue is not fixed. Thanks for the detailed clarification! I'm using keyworded kernel and headers, and frankly haven't noticed the warning, thus haven't looked deep enough into this issue. If QA is a blocker for getting this fixed, can we not patch in a cpupower_is_cpu_online prototype in mate-applets? The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=807be3e849bf1ec12d52cd65cf03c339c1a59e53 commit 807be3e849bf1ec12d52cd65cf03c339c1a59e53 Author: Mike Gilbert <floppym@gentoo.org> AuthorDate: 2017-09-12 19:12:20 +0000 Commit: Mike Gilbert <floppym@gentoo.org> CommitDate: 2017-09-12 19:13:29 +0000 sys-power/cpupower: install cpupower.h Bug: https://bugs.gentoo.org/593470 Package-Manager: Portage-2.3.6_p60, Repoman-2.3.3_p38 sys-power/cpupower/{cpupower-4.9.0.ebuild => cpupower-4.9.0-r1.ebuild} | 1 + 1 file changed, 1 insertion(+)} The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/sync/gentoo.git/commit/?id=807be3e849bf1ec12d52cd65cf03c339c1a59e53 commit 807be3e849bf1ec12d52cd65cf03c339c1a59e53 Author: Mike Gilbert <floppym@gentoo.org> AuthorDate: 2017-09-12 19:12:20 +0000 Commit: Mike Gilbert <floppym@gentoo.org> CommitDate: 2017-09-12 19:13:29 +0000 sys-power/cpupower: install cpupower.h Bug: https://bugs.gentoo.org/593470 Package-Manager: Portage-2.3.6_p60, Repoman-2.3.3_p38 sys-power/cpupower/{cpupower-4.9.0.ebuild => cpupower-4.9.0-r1.ebuild} | 1 + 1 file changed, 1 insertion(+)} (In reply to Joakim Tjernlund from comment #15) > If QA is a blocker for getting this fixed, can we not patch in a > cpupower_is_cpu_online prototype in mate-applets? Did some local overlay testing (following the recent bump for sys-power/cpupower-4.9.0-r1 which has a commit message referencing this bug, #593470) This particular -r1 version still errors for me, though now the build terminates on: undefined reference to `cpufreq_cpu_exists' (as opposed to cpupower_is_cpu_online) Unsure if the upstream mate devs test against a particular [older] kernel version or [patched] distro kernel, and/or certain cpupower version(s), etc. but I feel these regressions are because of the API changes. That is to say - this sort of build issue likely stems from cpupower internals having updated for a newer kernel API, as that interface has been changed by upstream kernel.org releases in the past year or so. (In reply to kuzetsa from comment #18) > (In reply to Joakim Tjernlund from comment #15) > > If QA is a blocker for getting this fixed, can we not patch in a > > cpupower_is_cpu_online prototype in mate-applets? > > Did some local overlay testing (following the recent bump for > sys-power/cpupower-4.9.0-r1 which has a commit message referencing this bug, > #593470) > > This particular -r1 version still errors for me, though now the build > terminates on: > > undefined reference to `cpufreq_cpu_exists' > > (as opposed to cpupower_is_cpu_online) > > Unsure if the upstream mate devs test against a particular [older] kernel > version or [patched] distro kernel, and/or certain cpupower version(s), etc. > but I feel these regressions are because of the API changes. > > That is to say - this sort of build issue likely stems from cpupower > internals having updated for a newer kernel API, as that interface has been > changed by upstream kernel.org releases in the past year or so. I bumped to cpupower-4.13.0 and it seems to be OK. It's in the tree. The kernel version check in the sources is still erroneous. It should be checking for the relevant userspace API functions, not the kernel version. NP-Hardass said he would work on an autoconf check. The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/proj/gentoo-mate.git/commit/?id=6bfa389a9bda535026178ba0e61f804bddb2505a commit 6bfa389a9bda535026178ba0e61f804bddb2505a Author: NP-Hardass <NP-Hardass@gentoo.org> AuthorDate: 2017-09-22 20:58:54 +0000 Commit: NP-Hardass <NP-Hardass@gentoo.org> CommitDate: 2017-09-22 20:58:54 +0000 mate-base/mate-applets: Fix building with >cpupower-4.9 Bug: https://bugs.gentoo.org/593470 Package-Manager: Portage-2.3.10, Repoman-2.3.3 .../files/mate-applets-1.10.4-cpupower-4.7.patch | 46 +++++++++ .../files/mate-applets-1.14.1-cpupower-4.7.patch | 44 +++++++++ .../mate-applets/mate-applets-1.12.1-r5.ebuild | 102 ++++++++++++++++++++ .../mate-applets/mate-applets-1.14.1-r2.ebuild | 106 +++++++++++++++++++++ .../mate-applets/mate-applets-1.16.0-r1.ebuild | 105 ++++++++++++++++++++ .../mate-applets/mate-applets-1.18.1-r1.ebuild | 96 +++++++++++++++++++ mate-base/mate-applets/mate-applets-9999.ebuild | 5 +- 7 files changed, 501 insertions(+), 3 deletions(-)} The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/proj/gentoo-mate.git/commit/?id=ec7e48d451844a3663272c0a2ba7dd83ec453593 commit ec7e48d451844a3663272c0a2ba7dd83ec453593 Author: NP-Hardass <NP-Hardass@gentoo.org> AuthorDate: 2017-09-22 21:17:37 +0000 Commit: NP-Hardass <NP-Hardass@gentoo.org> CommitDate: 2017-09-22 21:17:37 +0000 mate-base/mate-applets: Fix cpupower patch for 1.16+ Upstream removed a space, rendering old patch broken Bug: https://bugs.gentoo.org/593470 Package-Manager: Portage-2.3.10, Repoman-2.3.3 .../files/mate-applets-1.16.0-cpupower-4.7.patch | 44 ++++++++++++++++++++++ .../mate-applets/mate-applets-1.16.0-r1.ebuild | 2 +- .../mate-applets/mate-applets-1.18.1-r1.ebuild | 2 +- mate-base/mate-applets/mate-applets-9999.ebuild | 2 +- 4 files changed, 47 insertions(+), 3 deletions(-)} The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=53fd275242b53169b20f487efd38698e54e9c32c commit 53fd275242b53169b20f487efd38698e54e9c32c Author: NP-Hardass <NP-Hardass@gentoo.org> AuthorDate: 2017-09-22 21:26:40 +0000 Commit: NP-Hardass <NP-Hardass@gentoo.org> CommitDate: 2017-09-22 21:26:56 +0000 mate-base/mate-applets: Fix compilation with >=cpupower-4.7 Closes: https://bugs.gentoo.org/593470 Package-Manager: Portage-2.3.10, Repoman-2.3.3 .../files/mate-applets-1.10.4-cpupower-4.7.patch | 46 +++++++++ .../files/mate-applets-1.14.1-cpupower-4.7.patch | 44 +++++++++ .../files/mate-applets-1.16.0-cpupower-4.7.patch | 44 +++++++++ .../mate-applets/mate-applets-1.12.1-r5.ebuild | 102 ++++++++++++++++++++ .../mate-applets/mate-applets-1.14.1-r2.ebuild | 106 +++++++++++++++++++++ .../mate-applets/mate-applets-1.16.0-r1.ebuild | 105 ++++++++++++++++++++ .../mate-applets/mate-applets-1.18.1-r1.ebuild | 96 +++++++++++++++++++ 7 files changed, 543 insertions(+) |