[Webkit-unassigned] [Bug 72589] [GTK] Expose accessibility hierarchy in WebKit2 to ATK/AT-SPI based ATs

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=72589





--- Comment #22 from Carlos Garcia Campos <cgarcia at igalia.com>  2012-01-09 08:47:32 PST ---
(In reply to comment #21)
> (In reply to comment #20)
> > (In reply to comment #18)
> > 
> > > Because of the reasons exposed (and sorry for not bringing this last -very important- point up before), I think it's better to keep doing what I did, perhaps adding some extra checks as you suggest, and reviewing the code not to leak anything.
> > > 
> > > What do you think?
> > 
> > I see, you don't actually want to block when calling the process but when reading from stdout, and kill the process after a timeout. In that case, do not set the descriptor as non-blocking, no?
> 
> Argh! I explain poorly, definitely :-)

I understood it :-)

> I both want and don't want to block, beacuse of differente reasons, and at different points:
> 
>   * I don't wan't to block (so don't want to use g_spawn_sync()) because, if everything goes fine (server running and loading the test HTML properly), I want the server to keep running as a child process until I don't need it anymore. And this could mean keeping it alive and running while executing one or more tests.

ok, g_spawn_async_with_pipes() then.

>   * I want to block exactly after spawning the child process with spawn_async and before exiting the runTestServer() function, because I don't want any test to start running before being sure the server is ready, which is what I know through the "OK" message.

That's what I meant in my previous comment, you want to block when reading stdout. Do not set the file descriptor as non-blocking and use read or even g_io_channel_read_chars().

>   * I can't use GIO channels and the add_watch() thing, because I can't rely on the main loop to handle the response from the child process, since that would mean exiting runTestServer() too soon (before handling the "OK" message). This is related to the previous point (why I want to block there).

Yes, that's why I proposed to use a blocking file descriptor, you don't need to use watchers, because read will block.

>   * I need the non-blocking thing in the file descriptor to implement the timeout thing so, in case something goes wrong in the child process, we make sure to kill it after some time, so the test doesn't hang.

For the timeout thing you can use the thread killer I proposed. 

> Hope this time I explained better :-)

Yes :-) If you think it's better to use loop + non-blocking read + sleep, just ignore my proposal and fix your current one.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.