I am author of that code, so if it's causing flaky failures I'm obviously not happy.
So here is my explanation: there is a race. One running Chrome process has the socket open. When the second browser process starts, it also opens the socket. And in the test we're seeing a situation when both have it open. I think it might be possible for some tests where we tear down and start up the browser again in one test. Why do we assert in the test utility function? Because it is supposed to return the PID of the browser process holding the socket. When we have two, we can't tell which holds the socket, and which is just starting up and connecting to it (at least using current tools). However, the failure in the linked log doesn't look like that. If there was a leftover ui_tests binary, it should use different home directory (I think). I haven't seen this failing on the bots before, so if this happened only recently, it may be worth it to investigate the issue directly on the bots (higher chance to reproduce). On Sun, Dec 13, 2009 at 19:46, Drew Wilson <atwil...@chromium.org> wrote: > I'm seeing errors like this sporadically on the buildbots: > [15369:15369:1212/155951:124047683694:FATAL:/b/slave/chromium-rel-linux/build/src/chrome/test/chrome_process_util_linux.cc(33)] Expected exactly 1 process to > have socket open: 15770 16033 > I'm not really certain from code inspection what's going on here - does it > mean that there's a hung ui_tests on one of the builders or something? I > recently re-enabled some worker tests on linux, and I'm not sure whether > this is fallout from that or some other issue. > Anyone have any insights? > -atw > P.S. Example full > stdio: http://build.chromium.org/buildbot/waterfall/builders/Chromium%20Linux/builds/8455/steps/ui_tests/logs/stdio > > -- > Chromium Developers mailing list: chromium-dev@googlegroups.com > View archives, change email options, or unsubscribe: > http://groups.google.com/group/chromium-dev -- Chromium Developers mailing list: chromium-dev@googlegroups.com View archives, change email options, or unsubscribe: http://groups.google.com/group/chromium-dev
