Bug 134109 – Increase priority on SharedTimer source

• Status: REOPENED
• Component: WebKitGTK

[Andre Moreira Magalhaes]

While running webkit on a Debian virtual machine with limited resources I stumbled upon an issue where JS setTimeout callbacks were not being properly invoked due to machine limitations and the usage of a low priority on setTimeout timers.

After discussing the issue with Gustavo (kov), we decided to increase the SharedTimer priority to use the default glib priority. I would like to start a discussion on whether this is the proper way to fix this issue.

Patch to come.

[Andre Moreira Magalhaes]

Created attachment 233433 [details]
Patch

[Andre Moreira Magalhaes]

Created attachment 233434 [details]
Patch with proper changelog

Oops, forgot to update changelog, new patch in.

[Gustavo Noronha (kov)]

Comment on attachment 233434 [details]
Patch with proper changelog

This would be a problem for webkit1, but is a proper fix for the webkit2 model IMO, since we do not compete with GTK+ mainloop sources anymore.

[WebKit Commit Bot]

Comment on attachment 233434 [details]
Patch with proper changelog

Clearing flags on attachment: 233434

Committed r170529: <http://trac.webkit.org/changeset/170529>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Carlos Alberto Lopez Perez]

This caused a performance regression on the test Animation/balls:FrameRate. See: https://bugs.webkit.org/show_bug.cgi?id=134972

[Martin Robinson]

The priority of the timers is very sensitive and can negatively affect animations, so I recommend we roll this out until we understand the isseu.

[Andre Moreira Magalhaes]

(In reply to comment #7)
> The priority of the timers is very sensitive and can negatively affect animations, so I recommend we roll this out until we understand the isseu.

I agree here. I actually opened this bug more to start a discussion on the proper priorities to use for the various timers/sources as we had some issues with JS being starved and thus breaking some webapps that relied on setTimeout calls. This issue only happened in a very specific usecase where we had a browser running in a VM with limited resources.

I vote for reverting this but would like to see your opinion on what is the best approach to solve the issue with JS being starved when running on limited resources.

[Carlos Alberto Lopez Perez]

I reverted r170529 on r171343 <https://trac.webkit.org/r171343> because it caused a ~10% performance regression on the perf test Animation/balls: https://webkit.org/b/134972


Reopening this bug.