[PATCH libinput] timer: change the timer offset warning to a client bug

[Olivier Fourdan]

On 6 February 2018 at 05:21, Peter Hutterer <peter.hutterer at who-t.net>
wrote:

> This looks like a libinput bug (well, it does say "libinput bug" on the
> package) but it hasn't been that for a long time. The cause is almost
> always
> insufficient motivation to call libinput_dispatch() by the caller, or at
> least
> not doing it with the celerity libinput demands (and deserves, if I may say
> so).
>
> Up-, down- or side-grade it to a client bug, so the outrage can be
> directed elsewhere, preferably away from me. And add a section to the
> documentation, just in case someone actually reads this stuff.
>

Brilliant commit message :)


> Signed-off-by: Peter Hutterer <peter.hutterer at who-t.net>
> ---
>  doc/faqs.dox | 22 ++++++++++++++++++++++
>  src/timer.c  |  8 ++++----
>  2 files changed, 26 insertions(+), 4 deletions(-)
>
> diff --git a/doc/faqs.dox b/doc/faqs.dox
> index f0d85859..d5190926 100644
> --- a/doc/faqs.dox
> +++ b/doc/faqs.dox
> @@ -198,4 +198,26 @@ href="https://www.freedesktop.
> org/software/systemd/man/hwdb.html">udev
>  hwdb</a> or patches that include a change to the hwdb. See @ref hwdb for
>  details on the hwdb and how to modify it locally.
>
> + at section faq_timer_offset What causes the "timer offset negative" warning?
> +
> +libinput relies on the caller to call libinput_dispatch() whenever data is
> +available on the epoll-fd. Doing so will process the state of all devices
> +and can trigger some timers to be set (e.g. palm detection, tap-to-click,
> +disable-while-typing, etc.). Internally, libinput's time offsets are
> always
> +based on the event time of the triggering event.
> +
> +For example, a touch event with time T may trigger a timer for the time T
> +
> +180ms. When setting a timer, libinput checks the wall clock time to ensure
> +that this time T + offset is still in the future. If not, the warning is
> +logged.
> +
> +When this warning appears, it simply means that too much time has passed
> +between the event occuring (and the epoll-fd triggering) and the current
>

My spell-checker claims 2 “r” in “occurring”


> +time. In almost all cases this is an indication of the caller being
> +overloaded and not handling events as speedily as required.
> +
> +The warning has no immediate effect on libinput's behavior but some of the
> +functionality that relies on the timer may be impeded (e.g. palms are not
> +detected as they should be).
> +
>  */
> diff --git a/src/timer.c b/src/timer.c
> index f8de6662..584673a0 100644
> --- a/src/timer.c
> +++ b/src/timer.c
> @@ -86,10 +86,10 @@ libinput_timer_set_flags(struct libinput_timer *timer,
>         uint64_t now = libinput_now(timer->libinput);
>         if (expire < now) {
>                 if ((flags & TIMER_FLAG_ALLOW_NEGATIVE) == 0)
> -                       log_bug_libinput(timer->libinput,
> -                                        "timer %s: offset negative
> (-%dms)\n",
> -                                        timer->timer_name,
> -                                        us2ms(now - expire));
> +                       log_bug_client(timer->libinput,
> +                                      "timer %s: offset negative
> (-%dms)\n",
> +                                      timer->timer_name,
> +                                      us2ms(now - expire));
>         } else if ((expire - now) > ms2us(5000)) {
>                 log_bug_libinput(timer->libinput,
>                          "timer %s: offset more than 5s, now %d expire
> %d\n",
> --
> 2.14.3
>

>From someone who dealt with such a reported issue with gnome-shell
recently, it looks good to me!

With the missing “r” in the doc section, it's:

Reviewed-by: Olivier Fourdan <ofourdan at redhat.com>

Cheers,
Olivier
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/wayland-devel/attachments/20180206/28deefc7/attachment.html>