Bug 139921 – [iOS] Log better using FeatureCounter why PageCache is failing due to pruned resources

• Component: Page Loading
• Status: RESOLVED FIXED

[Chris Dumez]

Log better using FeatureCounter why PageCache is failing due to pruned resources. The reason can be:
- Page cache reached maximum capacity
- Memory pressure
- Process suspended (WK2)

Radar: <rdar://problem/19255690>

[Chris Dumez]

Created attachment 243707 [details]
Patch

[Chris Dumez]

Created attachment 243708 [details]
Patch

[Chris Dumez]

Created attachment 243710 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 243710 [details]
Patch

Clearing flags on attachment: 243710

Committed r177712: <http://trac.webkit.org/changeset/177712>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Alexey Proskuryakov]

This made a regression test flakily assert:

http://webkit-test-results.appspot.com/dashboards/flakiness_dashboard.html#showAllRuns=true&tests=fast%2Floader%2Fcrash-copying-backforwardlist.html

0   com.apple.JavaScriptCore      	0x0000000106ce8b8a WTFCrash + 42 (Assertions.cpp:321)
1   com.apple.WebCore             	0x000000010cb57a51 WebCore::pruningReasonToFeatureCounterKey(WebCore::PruningReason) + 193 (PageCache.cpp:451)
2   com.apple.WebCore             	0x000000010cb57ba2 WebCore::PageCache::get(WebCore::HistoryItem*, WebCore::Page*) + 306 (PageCache.cpp:512)
3   com.apple.WebCore             	0x000000010bee4164 WebCore::FrameLoader::loadDifferentDocumentItem(WebCore::HistoryItem*, WebCore::FrameLoadType, WebCore::FrameLoader::FormSubmissionCacheLoadPolicy) + 116 (FrameLoader.cpp:3178)
4   com.apple.WebCore             	0x000000010beee676 WebCore::FrameLoader::loadItem(WebCore::HistoryItem*, WebCore::FrameLoadType) + 166 (FrameLoader.cpp:3271)
5   com.apple.WebCore             	0x000000010c00183d WebCore::HistoryController::recursiveGoToItem(WebCore::HistoryItem*, WebCore::HistoryItem*, WebCore::FrameLoadType) + 141 (HistoryController.cpp:740)
6   com.apple.WebCore             	0x000000010c0015a5 WebCore::HistoryController::goToItem(WebCore::HistoryItem*, WebCore::FrameLoadType) + 405 (HistoryController.cpp:301)
7   com.apple.WebCore             	0x000000010cb45f99 WebCore::Page::goToItem(WebCore::HistoryItem*, WebCore::FrameLoadType) + 201 (Page.cpp:446)

[WebKit Commit Bot]

Re-opened since this is blocked by bug 139944

[Chris Dumez]

Ok, the problem seems to be that HistoryItem::m_prunedReason is not correctly initialized in the HistoryItem copy constructor.

[Chris Dumez]

Created attachment 243748 [details]
Patch

[Chris Dumez]

Comment on attachment 243748 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=243748&action=review

> Source/WebCore/history/HistoryItem.cpp:165
> +    , m_pruningReason(PruningReason::None)

This is what was missing in the previous patch that landed. I missed it because m_next / m_prev pointers weren't initialized here either. I am now initializing m_next / m_prev here as well as there is no guarantee those will be initialized to nullptr.

[Alexey Proskuryakov]

rs=me on the additional changes. I think that this should be landed under original Gavin's review.

[Chris Dumez]

Created attachment 243749 [details]
Patch

[WebKit Commit Bot]

Comment on attachment 243749 [details]
Patch

Clearing flags on attachment: 243749

Committed r177738: <http://trac.webkit.org/changeset/177738>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.