[Webkit-unassigned] [Bug 161481] [GTK] Use GTestDBus instead of dbus-launch in WebKitTestBus.cpp

bugzilla-daemon at webkit.org bugzilla-daemon at webkit.org
Wed Nov 2 05:45:52 PDT 2016


--- Comment #12 from Michael Catanzaro <mcatanzaro at igalia.com> ---
(In reply to comment #10)
> Why? Could you show a bt of one of such crashes? Because our current code is
> calling setenv in the same place, this patch is only replacing the code that
> GTestDBus with GTestDBus.

Here's one example of such a crash, with a backtrace:


We had a second case in some important GNOME component recently, but I honestly don't remember where and can't find it now. And we have third one in Endless's private bug tracker right now.

> > We need to be careful to only call
> > such functions (a) at the very top of main(),
> Which is exactly what happens. This is not public API, this is an internal
> helper class designed to be used by tests that should do this at the very
> beginning of beforeAll():
> bus = new WebKitTestBus();
> if (!bus->run())
>     return;
> And beforeAll() is called by main() right after setting up the environment
> and registering gresources. There isn't any other code before that, nor any
> other secondary thread created.

OK, then it should be OK, but it is so easy to misuse that I would prefer that you add an assertion to guarantee this. Unfortunately the best example I found requires reading /proc:


but it could be done inside #ifndef NDEBUG. Then if that check passes, we can be confident that GLib isn't creating any secondary thread somewhere before the setenv.

You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.webkit.org/pipermail/webkit-unassigned/attachments/20161102/9bc8c487/attachment.html>

More information about the webkit-unassigned mailing list