[Bug 110246] [CI][SHARDS] Random tests - incomplete - empty stdout/stderr/dmesg - watchdog0: watchdog did not stop!

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=110246

--- Comment #42 from Arek Hiler <arkadiusz.hiler at intel.com> ---
This is a pretty wild ride so far. There are a couple of things that are
happening here:

1. we may be closing watchdogs "twice", once by calling it explicitly on a
regular exit code path, and the second time from atexit() handler

since all those path are graceful(-ish), which is good because we close the
watchdogs at the end of execution phase before results' intermediaries are
parsed and compiled into .json

2. we should register signal handlers for SIGTERM, SIGHUP, SIGINT, SIGQUIT
while setting up watchdogs, as there is small window of opportunity between
each subtest execution that has not masked the signals. So if the signal is
timed just right we won't close the dogs.

execute_next_entry() is called for each "entry" (usually subtests) and does the
masking and unmasking.

3. we should increase runner's verbosity level in CI to get more data on
watchdog handling

working on patches to address some of the points here

-- 
You are receiving this mail because:
You are the assignee for the bug.
You are on the CC list for the bug.
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20190709/75e4d974/attachment.html>