Paul Chaignon's GSoC status report - #2 of 12

[Paul Chaignon]

Hi all,

I've focused on the status-qualifier patchset this week as I've had some
difficulties with tests in multi-threading contexts.

Accomplishments:
- Finished to code the cases for unavailable, exited, and detached
  statuses.
- Fixed many bugs in the implementation: exiting syscalls printed out even
  with -estatus=none, segfault when child thread execves leader thread (I
  had not adapted maybe_switch_tcbs), execve returning always printed (I
  needed to reopen a memstream in maybe_switch_tcbs), etc.
- Wrote a status-all test.  The status=all case is straightforward, but I
  thought it might be good to test the case where we're passing the full
  set (i.e., status=unavailable,exited,detached,failed,successful).
- Set status to default to unavailable, exited, and detached sequentially
  to see which tests were failing in each case and if the fails made
  sense.  That helped me find several bugs and identify candidate
  behaviors to write new test cases.
- Wrote test cases for status=exited and status=none based on the
  threads-execve test.  status=none currently only prints the '+++' exit
  messages.  Both tests give me the expected output under strace when
  launched manually but the output corresponding to the child process is
  missing when launched as part of the test suite.  I'm debugging this
  right now.

Priorities:
- Finish status-exited and status-none tests.
- Find a way to test status-detached.
- Send status-qualifier patchset.
- Go back to working on seccomp filtering!

Paul