@proxy-maintainers : Please update the tree with the ebuild for the 3.9.5 release of the ck-sources and the ck-sources-3.9-virt_cpu_accounting_gen-aCOSwt_P7.patch file. Note : Despite our advices and the discouraging help text documenting the "Full dynticks CPU time accounting" kernel option, it happens that an always increasing number of users selects it together with the "BFS cpu scheduler" This will incidentally break the build of bfs.c Because it makes actually very little sense (if any) to set the "Full dynticks CPU time accounting" kernel option, at least on a ck-sources-3.9 with the "BFS cpu scheduler", removing this option from the kernel configuration menus will ease users' experience of the ck-sources and lower the support workload to absolutely no functional expense. This is the purpose of the attached patch. Reproducible: Always
Created attachment 350620 [details] ck-sources-3.9.5.ebuild Ebuild for the 3.9.5 release. Based on : - Sources for the linux 3.9 branch - genpatches rev 9 - 3.9-ck1 patchset including bfs-430
Created attachment 350622 [details, diff] files/ck-sources-3.9-virt_cpu_accounting_gen-aCOSwt_P7.patch Patch modifying init/Kconfig in order to make VIRT_CPU_ACCOUNTING_GEN depends on !SCHED_BFS
Comment on attachment 350620 [details] ck-sources-3.9.5.ebuild --- ck-sources-3.9.2.ebuild 2013-05-15 20:58:00.000000000 +0200 +++ - 2013-06-10 14:14:23.263093266 +0200 @@ -1,6 +1,6 @@ # Copyright 1999-2013 Gentoo Foundation # Distributed under the terms of the GNU General Public License v2 -# $Header: /var/cvsroot/gentoo-x86/sys-kernel/ck-sources/ck-sources-3.9.2.ebuild,v 1.1 2013/05/15 18:58:00 hwoarang Exp $ +# $Header: $ EAPI="5" ETYPE="sources" @@ -11,7 +11,7 @@ http://users.on.net/~ckolivas/kernel/" K_WANT_GENPATCHES="base extras" -K_GENPATCHES_VER="5" +K_GENPATCHES_VER="9" K_SECURITY_UNSUPPORTED="1" K_DEBLOB_AVAILABLE="1" @@ -62,7 +62,7 @@ #-- Local patches needed for the ck-patches to apply smoothly (3.4/3.5) ------- PRE_CK_FIX="" -POST_CK_FIX="" +POST_CK_FIX="${FILESDIR}/${PN}-3.9-virt_cpu_accounting_gen-aCOSwt_P7.patch" #--
If this symbol makes no sense for ck-sources, shouldn't it be part of the BFS patchset?
(In reply to Markos Chandras from comment #4) > If this symbol makes no sense for ck-sources, shouldn't it be part of the > BFS patchset? Sorry Markos, but I do not exactly understand what you are challenging there : - The fact that enabling the "Full dynticks CPU time accounting" kernel option makes very little sense, or, - The fact that my patch is not yet included in the latest ck patchset ?
(In reply to Eric F. GARIOUD from comment #5) > (In reply to Markos Chandras from comment #4) > > If this symbol makes no sense for ck-sources, shouldn't it be part of the > > BFS patchset? > Sorry Markos, but I do not exactly understand what you are challenging there > : > - The fact that enabling the "Full dynticks CPU time accounting" kernel > option makes very little sense, or, > - The fact that my patch is not yet included in the latest ck patchset ? I mean, if there is need to apply a separate patch to hide that symbol, shouldn't the patch you attached here be included in the BFS patchset? If BFS does not support it then it should be in the BFS patchset. Otherwise, I wouldn't care much, let users enable it if they want to
(In reply to Markos Chandras from comment #6) > (In reply to Eric F. GARIOUD from comment #5) > > (In reply to Markos Chandras from comment #4) > > > If this symbol makes no sense for ck-sources, shouldn't it be part of the > > > BFS patchset? > > Sorry Markos, but I do not exactly understand what you are challenging there > > : > > - The fact that enabling the "Full dynticks CPU time accounting" kernel > > option makes very little sense, or, > > - The fact that my patch is not yet included in the latest ck patchset ? > If BFS does not support it... This is a fact, build of bfs.c is broken if (CONFIG_VIRT_CPU_ACCOUNTING_GEN=y && CONFIG_SCHED_BFS=y). > then it should be in the BFS patchset. This is also my opinion, as far as the 3.9 branch is concerned. However, because you already wrote this as part of another bug, I know you are aware that the ck-sources "is a kernel patchset that has no upstream support" > I wouldn't care much, let users enable it if they want to Hmmm... I care a little more, mainly... because I am the one being mailedto when a build breaks... and, honestly, if I do not care much of users making their kernels run badly because of inappropriate configurations, I am forced to agree that a kernel should build irrespective of user's selectable options. This being said, I perfectly acknowledge that this patch is at first motivated by package maintenance comfort considerations. So... feel free to ignore and close WONTFIX.
(In reply to Eric F. GARIOUD from comment #7) > (In reply to Markos Chandras from comment #6) > > (In reply to Eric F. GARIOUD from comment #5) > > > (In reply to Markos Chandras from comment #4) > > > > If this symbol makes no sense for ck-sources, shouldn't it be part of the > > > > BFS patchset? > > > Sorry Markos, but I do not exactly understand what you are challenging there > > > : > > > - The fact that enabling the "Full dynticks CPU time accounting" kernel > > > option makes very little sense, or, > > > - The fact that my patch is not yet included in the latest ck patchset ? > > > If BFS does not support it... > This is a fact, build of bfs.c is broken if > (CONFIG_VIRT_CPU_ACCOUNTING_GEN=y && CONFIG_SCHED_BFS=y). > > > then it should be in the BFS patchset. > This is also my opinion, as far as the 3.9 branch is concerned. > > However, because you already wrote this as part of another bug, I know you > are aware that the ck-sources "is a kernel patchset that has no upstream > support" It is but what I am saying is that this patch should be included in the BFS patchset because it's related to the BFS scheduler. This will make it easier to maintain during version bumps. That's my opinion. Anyway, i don't mind i just think it might be a pain for you to maintain it in the long-term future.
Created attachment 351750 [details] ck-sources-3.9.7.ebuild Ebuild for the 3.9.7 release. Based on : - Sources for the linux 3.9 branch - genpatches rev 12 - 3.9-ck1 patchset including bfs-430 This being obviously to be preferred to the now obsolete 3.9.5, original topic of this bug. Note : The BFQ I/O scheduler is part of the ck-sources from this release.
+ 25 Jun 2013; Sergey Popov <pinkbyte@gentoo.org> + +files/ck-sources-3.9-virt_cpu_accounting_gen-aCOSwt_P7.patch, + +ck-sources-3.9.7.ebuild: + Version bump, wrt bug #472842, thanks to Eric F. GARIOUD <eric-f.garioud AT + wanadoo.fr>
Thank you, Sergey.