Recently I started encountering the following build errors in the automated builds: ERROR: maintenance-image-5.1.9999-r0 do_rootfs: The sstate manifest for task 'libarchive-native:populate_sysroot' (multilib variant '') could not be found. The pkgarchs considered were: x86_64, x86_64_fedora-39. But none of these manifests exists: /yocto/scarthgap/build/tmp/sstate-control/manifest-x86_64-libarchive-native.populate_sysroot /yocto/scarthgap/build/tmp/sstate-control/manifest-x86_64_fedora-39-libarchive-native.populate_sysroot and ERROR: linux-yocto-6.6.84+git-r0 do_prepare_recipe_sysroot: The sstate manifest for task 'xz-native:populate_sysroot' (multilib variant '') could not be found. The pkgarchs considered were: x86_64, x86_64_fedora-39. But none of these manifests exists: /yocto/scarthgap/build/tmp/sstate-control/manifest-x86_64-xz-native.populate_sysroot /yocto/scarthgap/build/tmp/sstate-control/manifest-x86_64_fedora-39-xz-native.populate_sysroot These errors started to appear after poky patch level updates (from 5.0.8 to 5.0.9 and .10). The errors can be reproduced/seem to happen, by switching between these versions. - build an image/recipe with one poky version - build same image/recipe with another poky version - build same image/recipe with 1. poky version, ensure any recipe needs rebuilding that requires now a changed version of the native recipe. I suspected some kind of broken sstate and completely cleared everything: build, tmp, sstate (except downloads). After above steps, the problem reappeared after a couple of builds. Observations: - all 'clean' or updated builds run without errors. - not directly tied to poky native recipes, the same was seen with up/downgrade of qtbase-native - the failing native recipes are exactly the ones with updated/changed versions between poky patch-level tags (seen for libarchive, xz). - do_cleansstate xxx-native fixes it for the next build.
Add Yoann who is maintaining the stable releases now. Mattias, Are you still seeing this problem? Are you able to reproduce it with just oe-core ?
Bump to 5.3.4. The description contains a reproducer albeit with a bit of ambiguity. We just don't have anyone signed up to work on the bug. Matthias, is this still a problem for you ? If so, can you either work on debugging it or add share the exact steps needed to reproduce the bug? Hopefully those steps only involve poky.
The issue is no longer a problem. I found, that the issue came from reusing the build and expecially tmp directories for different machine builds (both using the same tune). Now the issue does no longer seem to appear after providing different directories to the poky init script (build-$MACHINE). sstate is still shared. In the meantime a lots of yocto/poky updates (latest to 5.0.17) with several rebuilds of native recipes (up and downgrade builds) never seem to trigger the described behaviour.
Hi Matthias, Thanks for the update. I'll close the issue. We do appreciate the bug report and may have more people available to work on such issues soon so if it comes back, please re-open this defect. We are working to ensure that sstate always "just works". ../Randy