Hi Ernast A few weeks ago I sent a patch series that tries to solve this problem. But refactoring the kernel and especially the kernel-fitimage and u-boot classes is not easy. My patches solve the problem with rebuilding, but they are really hard to review, so they were rejected here: https://lists.openembedded.org/g/openembedded-core/topic/107982760#msg203695 .
The next idea is to split fitImage into separate recipes, hoping that this will simplify the situation and allow us to better utilize the sstate-cache. So far I have not tackled this yet. Thanks for bringing this up. Regards, Adrian Am Di., 1. Okt. 2024 um 15:29 Uhr schrieb Ernast via lists.yoctoproject.org <ersevs=gmail....@lists.yoctoproject.org>: > Hello, > > I have been looking into improving the build time of my yocto build and > following some of the recommendations regarding debugging long build times, > I have traced the issue back to dynamic data in my rootfs which causes > misses in the sstate cache. I believe the dynamic data causes the following > rebuild chain: do_rootfs -> kernel do_compile -> do_bundle_initramfs. > > I understand why do_rootfs/do_bundle_initramfs must be rerun but in this > particular case there shouldn't be a need to recompile the kernel. As far > as I can tell, this is only happening because the final initramfs (which > needs rebuilt) and the bzImage are part of the same sstate artifact (which > is invalidated as a result of the dynamic data in the rootfs). I'm > wondering if there is an opportunity to improve this and keep the kernel > from recompiling each time only the rootfs changes. > > Unfortunately the dynamic data in the rootfs is a hard requirement for me > and cannot be removed. > > Thanks, > > Ernast > > > >
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#63921): https://lists.yoctoproject.org/g/yocto/message/63921 Mute This Topic: https://lists.yoctoproject.org/mt/108758712/21656 Group Owner: yocto+ow...@lists.yoctoproject.org Unsubscribe: https://lists.yoctoproject.org/g/yocto/unsub [arch...@mail-archive.com] -=-=-=-=-=-=-=-=-=-=-=-
