Closed myo73 closed 5 years ago
The observed problem is the side effect of commit b92589c - vmlinuz and initrd are overwritten for the same machine by subsequent build of other distro.
This explains why no problems were detected by "fast" smoke_test - only stretch distros for different QEMU target architectures were built.
Also it's clear why the problem was detected for qemuarm and i386 - example module was built for 4.9 kernel shipped with stretch.
This 4.9 kernel was overwritten later by 4.19 kernel by buster build under deploy/images/qemuarm and deploy/images/qemui386 - VM with stretch rootfs and 4.19 kernel starts OK, but, naturally, it fails to load the example module built for 4.9 kernel.
Starting from overnight build #291 of "standard" CI in the 'next' branch http://isar-build.org:8080/job/isar_next/291 permanent failures of "Check example module" test for qemuarm stretch and qemui386 stretch are observed.
This happened after the merge of v2 patch series "implement ROOTFS postprocess commands" by Claudius Heine on 16th of April, 2019.
The interesting point that this happens for "standard" CI, but for "fast" CI example module test for qemuarm-stretch passes OK: http://isar-build.org:8080/job/isar_next_fast/1/console
From the first glance the difference (compared with master's tree) is with kernel version - previously that was 4.9.0-9 - now it became 4.19.0.4.
This blocks now the sync of 'master' branch with the 'next' branch.