Bug 98848
| Summary: | robohornet gets very slow if the page is backgrounded | ||
|---|---|---|---|
| Product: | WebKit | Reporter: | Kiran Muppala <cmuppala> |
| Component: | DOM | Assignee: | Kiran Muppala <cmuppala> |
| Status: | NEW | ||
| Severity: | Minor | CC: | ap, fishd, jamesr, kbr, mjs, simon.fraser, tonikitoo |
| Priority: | P5 | ||
| Version: | 528+ (Nightly build) | ||
| Hardware: | Mac (Intel) | ||
| OS: | Unspecified | ||
Kiran Muppala
https://bugs.webkit.org/show_bug.cgi?id=98474 introduced throttling of DOM timers on hidden pages by aligning them to 1 second intervals. Kenneth Russell, commented in that bug that a known issue with timer throttling in chromium is robohornet benchmark (www.robohornet.org) slowing down if put on a background tab, even if it has popup windows which are in the foreground. http://code.google.com/p/chromium/issues/detail?id=152077.
Testing on mac, showed that hidden page timer alignment also exhibits the same behavior. Filing this bug to figure out how to address this issue.
| Attachments | ||
|---|---|---|
| Add attachment proposed patch, testcase, etc. |
Maciej Stachowiak
Is this actually a bug? I'd expect benchmarks that rely on timers to run slower when in a background tab, with any form of background tab timer throttling.
We could add a special case to avoid throttling background (or otherwise non-visible) pages with visible popups, but is there a reason that is actually important?
Kiran Muppala
(In reply to comment #1)
> Is this actually a bug? I'd expect benchmarks that rely on timers to run slower when in a background tab, with any form of background tab timer throttling.
>
> We could add a special case to avoid throttling background (or otherwise non-visible) pages with visible popups, but is there a reason that is actually important?
No, I didn't think this is important. In fact I wasn't even sure if this needed to be fixed. I just filed the bug to not loose track of it. Lowered the priority of the bug.
Kenneth Russell
I suspect that the authors of the robohornet benchmark consider this more than a minor issue. CC'ing a couple.
Ojan Vafai
I think we should just change robohornet to listen to visibilitychange events and return an error to the user if the page is ever hidden.