Hi Time, > As indicated already, please consider this patch for the next version of > splashy. It replaces the sleep in the script, and only holds up init for > the time necessary (certainly less than 0.5 s and perhaps only 0.1 s) > (attached) Turns out the patch was easier than I expected :-)
>
> + /* register signal handlers if we are doing splashy boot */
> + if (g_ascii_strncasecmp (argv[1], "boot", 4) == 0) {
> + (void) signal (SIGCHLD, sig_exit_boot); /* this signal is
> sent when a child process exits */
> + (void) signal (SIGUSR2, sig_exit_boot); /* this is the
> signal sent by _splashy_child indicating directfb is fully initialised */
> +
> /* so the parent process can exit, letting init continue. We lock
> initramfs from */
> +
> /* being unmounted while the child process is still setting up
> directfb. */
> +
> /* some machines had a race condition causing splashy failure */
> +
> /* This fixes bug 505270 */
> + }
> +
You're messing up the indentation here, presumably because your not mixing
tabs and spaces properly (though I don't see any real tabs in the patch..)
> @@ -141,6 +177,8 @@
> {
> DEBUG_PRINT ("Calling splashy_child_start()");
> splashy_child_start ();
> +
> +
> }
> else if (g_ascii_strncasecmp (argv[1], "shutdown", 8) == 0)
> {
This is a pretty useless hunk :-)
> + /* Wait 1s for signal. There are two possible signals. */
> + /* The child may die (splashy failed), or the child process
> signals that directfb is setup. */
> + /* If the exit signal does not come in this time, then exit
> anyway */
> + /* the signal should come very quickly, so this approach
> saves most of a second */
> +
> + g_usleep(1000000); /* 1s */
> + g_printerr ("%s", "Error occured : splashy boot
> terminated; timed out waiting for splashy to initialise\n");
> + }
Perhaps it would be good to increase this sleep interval? On normal operation,
the time isn't reached anyway, and on slow computers / boot media it could
perhaps take long time. It would be annoying if this bug would be reintroduced
just because a particular machine or boot takes a bit longer. How about making
it 10s?
Also, I would write this as:
g_usleep(10*1000*1000)
It's a bit clearer IMHO.
Gr.
Matthijs
signature.asc
Description: Digital signature
