Bug 40072 – [Qt] QtTestBrowser random crashes are observed while using QtWebkit 2.0

• Component: New Bugs
• Status: RESOLVED DUPLICATE

[Alexandra Santos]

QtTestBrowser randomly crashes while loading pages.

Steps to reproduce:

1.- On n97, install Qt.sis (4.6.2), QtWebkit.sis (QtWebkit 2.0)and QtTestBrowser.
2.- Launch QtTestBrowser and start loading pages.

Actual results:

Random crashes are observed.

Expected results:

QtTestBrowser would not crash while loading pages.

NOTE: Please find attached crash logs.

[Alexandra Santos]

Created attachment 57667 [details]
QtTestBrowser crash logs

[Simon Hausmann]

Thank you Alexandra for the crash logs. Can you tell us which qtwebkit.sis file you were using?

Alternatively, can you provide us with the map file of QtWebKit, so that we can analyze the crash logs?

[Laszlo Gombos]

Setting it to P2 (from P3). It would be nice to confirm that this "random crashes" are _not_ as a result of an out-of-memory situation.

[Alexandra Santos]

I was using build 39, map files are located in the same folder. Please contact me if you need more details regarding this issue.

I'm able to load pages without any crash on latest WebKit 2.0 version with WrtLite browser.
Let me know if you need any more details.

[Simon Hausmann]

As far as I can see from the crash logs, these appear to be out-of-memory crashes.

[Jesus Sanchez-Palencia]

(In reply to comment #6)
> As far as I can see from the crash logs, these appear to be out-of-memory crashes.

Don't we have other open bugs for out-of-memory crashes already?
Should we keep this one open as well?

[ananth jasty]

Based on some testing I've found that the system is generally refusing to release memory and increasing its reserved size over time, until a high-water-mark is reached, at which point some pages are released. This process however, has diminishing returns, with less memory being released after each low memory event.

Assuming even that pages are being kept for efficiency, is there a way to analyse page utilization, and fragmentation?

I am hoping to check the most likely culprit for large memory leaks, the qpixmap system, to see if we are simply not freeing images often enough, but also, is it possible the high-water-mark is not triggering a cache-purge aggressively enough?

[Simon Hausmann]

Indeed, these crashes are out of memory crashes. Let's track out of memory on Symbian with the tests that QA runs with bug 40446.

*** This bug has been marked as a duplicate of bug 40446 ***