Hi Petr,
Given the realization about kernel log timestamps and partial log comparison with v2, I respun a final version dropping the dmesg --notime patch, fixed any rebase conflicts, and added a comment per your suggestion.
I copied all the ack and review tags from v2 since the patchset is unchanged otherwise. Hopefully this v3 minimizes any maintainer fiddling on your end.
I did iterate through the patches and verified that I could run each multiple times without the dmesg comparison getting confused.
Thanks,
-- Joe
v3:
- when modifying the dmesg comparision to select only new messages in patch 1, add a comment explaining the importance of timestamps to accurately pick from where the log left off at start_test [pmladek]
- since Petr determined that the timestamps were in fact very important to maintain for the dmesg / diff comparision, drop the patch which added --notime to dmesg invocations [pmladek]
- update the comparision regex filter for 'livepatch:' now that it's going to be prefixed by '[timestamp] ' and no longer at the start of the buffer line. This part of the log comparison should now be unmodified by the patchset.
Joe Lawrence (3): selftests/livepatch: Don't clear dmesg when running tests selftests/livepatch: refine dmesg 'taints' in dmesg comparison selftests/livepatch: add test delimiter to dmesg
tools/testing/selftests/livepatch/README | 16 +++--- .../testing/selftests/livepatch/functions.sh | 37 ++++++++++++- .../selftests/livepatch/test-callbacks.sh | 55 ++++--------------- .../selftests/livepatch/test-ftrace.sh | 4 +- .../selftests/livepatch/test-livepatch.sh | 12 +--- .../selftests/livepatch/test-shadow-vars.sh | 4 +- .../testing/selftests/livepatch/test-state.sh | 21 +++---- 7 files changed, 68 insertions(+), 81 deletions(-)