Control: tag -1 unreproducible help
Control: severity -1 important

Hello Lucas,

first, thanks for running these rebuild tests!

Lucas Nussbaum [2016-12-03  8:44 +0100]:
> > ERROR:tests/test-umockdev-run.c:1550:t_input_touchpad: assertion failed 
> > (_tmp52_ == ""): ("Unable to connect to X server\n" == "")
> > /bin/bash: line 1: 93517 Aborted                 TOP_BUILDDIR=. 
> > TOP_SRCDIR=. LD_LIBRARY_PATH=./.libs:$LD_LIBRARY_PATH GI_TYPELIB_PATH=. 
> > PATH=./src:$PATH G_SLICE=debug-blocks MALLOC_PERTURB_=85 MALLOC_CHECK_=3 
> > ./src/umockdev-wrapper $f

I tried to run a complete package build in a clean sbuild sid environment, and
also ran the test in a loop:

| $ for i in `seq 100`; do PATH=src:$PATH LD_LIBRARY_PATH=.libs 
src/umockdev-wrapper tests/test-umockdev-run -p 
/umockdev-run/integration/input-touchpad; done
| /umockdev-run/integration/input-touchpad: OK
| [...]

and I'm afraid I'm unable to reproduce this. Indeed this is the only test that
spawns an X.org server with the dummy driver:

  
https://anonscm.debian.org/cgit/collab-maint/umockdev.git/tree/tests/test-umockdev-run.vala#n443

so that it can run that under umockdev and verify that the input devices get
simulated correctly. This already waits until /tmp/.X11-unix/X5 exists, which
AFAIK is the official signal from X.org that it started up sufficiently to
allow clients to connect to it. This obviously worked, as otherwise the test
would show "SKIP: Xorg failed to start up..." (and not fail).

Can you reproduce this somehow (particularly with the above loop, in a built
tree)? If you can, it would be great if you could try and add

  Posix.sleep (2);
 
after the loop that polls for the socket, i. e. to line 492. If that helps,
then it's a race condition with X.org's startup, if it does not help then
the dummy Xorg server is somehow broken in that environment.

Since this is unreproducible and it doesn't seem to happen on the Debian
builders either, I downgrade to important. If you disagree, feel free to bump
it back, but I don't think this is a sufficient reason to kick it out of
testing.

Thanks,

Martin
-- 
Martin Pitt                        | http://www.piware.de
Ubuntu Developer (www.ubuntu.com)  | Debian Developer  (www.debian.org)

Reply via email to