upstart,libnih ftbfs on s390x with linux 4.4.0-21.37
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
upstart (Ubuntu) |
New
|
Undecided
|
Dimitri John Ledkov |
Bug Description
The latest uploads of upstart are failing to build on s390x, with lost fds causing the testsuite to fail to return (which should eventually lead to a timeout, but hasn't as of this writing):
https:/
https:/
upstart also fails to build on devac02, with the following test failures:
not ok 12 - ensure log written when directory created accessible with uid 0
wrong value for output, got unexpected (nil)
at tests/test_
/bin/bash: line 5: 31104 Aborted (core dumped) ${dir}$tst
FAIL: test_log
[...]
not ok 19 - with deletion of top-level directory
wrong value for source->watch, expected (nil) got 0x2aa43d27cb0
at tests/test_
Aborted (core dumped)
FAIL: test_conf_
The second of these failures definitely looks like a lost inotify event.
So I looked at libnih, and libnih's testsuite hangs indefinitely on devac02, after:
PASS: test_file 78 - with simple directory loop
This is the last test in test_file; the next test that it hangs on is test_watch. Running this test directly shows that it hangs after:
ok 17 - nih_watch_reader
strace shows the test_watch waiting indefinitely on a select, waiting to read the inotify fd.
All of this is on systems running linux 4.4.0-21.
The upstart testsuite was passing as recently as 4.4.0-18.
https:/
I don't know what in the kernel could have changed to break this, but that seems the most likely explanation. There was also a new glibc version since the last successful upstart build, 2.23-0ubuntu3, which includes an s390x-specific change, but that seems quite unlikely to impact inotify.
Changed in upstart (Ubuntu): | |
assignee: | nobody → Dimitri John Ledkov (xnox) |
Just to be sure, I tried downgrading libc6 in the build environment from 2.23-0ubuntu3 to 2.23-0ubuntu2. The problem is still reproducible.
And the build filesystem is ext4, which should be fairly inotify-safe.