**** BEGIN LOGGING AT Tue Jan 31 03:00:02 2017 Jan 31 21:19:41 bluelightning, omg...sightening of lumag Jan 31 21:19:55 hi ant_home Jan 31 21:20:03 where? Jan 31 21:20:08 #oe Jan 31 21:24:11 bluelightning: I think I'm understanding now how to manage the tasks Jan 31 21:24:41 it is not clear to me how to output the task dependencies Jan 31 21:25:04 atm I'm just looking at task.order and listtasks output Jan 31 21:26:12 I have to say it is not immediately logic how deleting some tasks impact others Jan 31 21:28:55 well, one major difference between noexec and deltask is that deltask will rip out the dependencies on that task as well, whereas noexec just makes the task do nothing Jan 31 21:29:41 i.e. if you have task A depending on task B that depends on task C, and you deltask B, task C will no longer get run as a dependency of task A Jan 31 21:30:16 (well, task C was never a dependency of task A directly, but it won't be in the chain anymore is what I mean, so it may never get run at all) Jan 31 21:34:35 I have tested that do_install[noexec] must be kept for kernels, otherwisw you don't see strip, sizecheck, staging Jan 31 21:34:58 it is also true you can readd these tasks... Jan 31 21:35:38 all the rest in our -kexecboot kernels can be deltask Jan 31 21:36:19 special case is for mips/vmlinuz/vmlinuz.bin handling, here a task must be added Jan 31 21:36:31 (as linux-yocto does) Jan 31 21:36:50 so why are you making this change? Jan 31 21:36:59 noexec isn't going anywhere... Jan 31 21:37:24 well, first was empty task, then noxec, finally (?) deltask ? Jan 31 21:37:37 how do we get rid of these tasks? Jan 31 21:38:05 you say finally - but I still don't get what's pushing this change? Jan 31 21:38:08 it is rather annoying that after defining PACKAGES = "" we still have to delete all packaging tasks... Jan 31 21:38:27 we have a class for that - inherit nopackage Jan 31 21:38:28 I see oe-core has a couple of commits replacing Jan 31 21:38:52 idea is to run less tasks Jan 31 21:38:59 actually, nopackages Jan 31 21:39:10 that does do deltask Jan 31 21:39:12 hm.. need to check that Jan 31 21:39:30 the advantage being that class gets updated if we change the packaging task structure Jan 31 21:40:49 see, the fundamental issue is that building two kernels is not as before Jan 31 21:41:23 so for our purposes we just need the artifacts in deploydir, not the packages Jan 31 21:42:01 ah, right Jan 31 21:42:10 and to not harm the shared kernel-source we don't nstall anything Jan 31 21:42:33 but it is still corrupted / racing during build Jan 31 21:42:58 i. e. build of linux-kexecboot fails if linux-handheld was not built before Jan 31 21:43:17 it tries to build it parallely and fails/races Jan 31 21:43:35 but this is a separate issue Jan 31 21:43:42 don't you just need a specific task dependency to fix that? Jan 31 21:43:45 for now, let just build kernel properly :) Jan 31 21:43:52 hm.. Jan 31 21:45:07 I think there is a new virtual/kernel: do_xx dep Jan 31 21:45:53 smthg like the one in image.bbclass Jan 31 21:45:56 do_build[depends] += "virtual/kernel:do_deploy" Jan 31 21:46:18 that's why I'm wondering how to visualize the task dependencies Jan 31 21:46:50 I'm slowly catching up with the changes to the classes Jan 31 21:46:53 :) Jan 31 21:51:01 bluelightning, anyway, once you push meta-handheld, let's discuss these two Jan 31 21:51:03 http://pastebin.com/U29LK5Gj Jan 31 21:52:54 ok, will try to do that tonight Jan 31 21:52:57 thanks Jan 31 21:53:19 so, if you see a way to remove all that packaging stuff...thanks Jan 31 21:53:55 I would strongly recommend inherit nopackages Jan 31 21:54:32 I can test it right now Jan 31 21:58:04 bluelightning, after this round of (build) fixes there are the staging fixe for sstate rebuilds... Jan 31 21:58:21 there I see light... Jan 31 21:59:00 I mean, with recipe specific sysroot the tricky klcc-cross mangling could be removed (I think) Jan 31 21:59:35 then we need to fix the simple recipes like zaurus updater just staging a file Jan 31 21:59:51 these fails from sstate as it was last week Jan 31 22:00:56 bluelightning, nopackages rulez Jan 31 22:01:09 let me retry... Jan 31 22:01:33 hmm, sstate rebuilds shouldn't be any different now with recipe specific sysroots Jan 31 22:01:53 well, actually that may not be the case Jan 31 22:03:33 in our case recipe A wraps-up the artifacts of B from staging dir Jan 31 22:06:04 ok, retested and wonderful Jan 31 22:06:05 # disable unneeded tasks Jan 31 22:06:06 do_install[noexec] = "1" Jan 31 22:06:06 deltask do_shared_workdir Jan 31 22:06:06 -deltask do_package Jan 31 22:06:06 -deltask do_package_qa Jan 31 22:06:07 -deltask do_packagedata Jan 31 22:06:09 -deltask do_package_deb Jan 31 22:06:11 -deltask do_package_ipk Jan 31 22:06:13 -deltask do_package_rpm Jan 31 22:06:15 -deltask do_package_tar Jan 31 22:06:17 -deltask do_package_write_deb Jan 31 22:06:19 -deltask do_package_write_ipk Jan 31 22:06:21 -deltask do_package_write_rpm Jan 31 22:06:23 -deltask do_package_write_tar Jan 31 22:06:27 deltask do_populate_sysroot Jan 31 22:06:29 + Jan 31 22:06:31 +inherit nopackages Jan 31 22:06:36 my bad for ignoring it up to now :/ Jan 31 22:18:29 bluelightning, I've added the last patches on the top Jan 31 22:18:31 https://github.com/LinuxPDA/meta-handheld/commits/master Jan 31 22:18:48 pls take your time, I'll be away till monday Jan 31 23:36:50 bluelightning, last news after pull of today Jan 31 23:36:54 | configure: error: C compiler cannot create executables Jan 31 23:37:05 heh, will see next week Jan 31 23:37:10 good night **** ENDING LOGGING AT Wed Feb 01 03:00:01 2017