Bug 173972 – Avoid copying ResourceLoadStatistics objects

• Status: RESOLVED
• Resolution: FIXED
• Component: WebCore Misc.

[Chris Dumez]

Avoid copying ResourceLoadStatistics objects given that they are big. Make the type move-only to avoid such mistakes in the future.

[Chris Dumez]

Created attachment 314139 [details]
Patch

[Chris Dumez]

Comment on attachment 314139 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=314139&action=review

> Source/WebCore/loader/ResourceLoadObserver.cpp:-155
> -        auto targetStatistics = m_store->ensureResourceStatisticsForPrimaryDomain(targetPrimaryDomain);

Note that I stopped calling ensureResourceStatisticsForPrimaryDomain() here. Instead, I merely construct a ResourceLoadStatistics...

> Source/WebCore/loader/ResourceLoadObserver.cpp:204
> +            m_store->setResourceStatisticsForPrimaryDomain(targetPrimaryDomain, WTFMove(targetStatistics));

... which gets inserted here.

This avoids copying the statistics object and I don't think the call to ensureResourceStatisticsForPrimaryDomain() was really required here.

[Geoffrey Garen]

Comment on attachment 314139 [details]
Patch

r=me

[WebKit Commit Bot]

Comment on attachment 314139 [details]
Patch

Clearing flags on attachment: 314139

Committed r218944: <http://trac.webkit.org/changeset/218944>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.

[Brent Fulgham]

Comment on attachment 314139 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=314139&action=review

>> Source/WebCore/loader/ResourceLoadObserver.cpp:-155
>> -        auto targetStatistics = m_store->ensureResourceStatisticsForPrimaryDomain(targetPrimaryDomain);
> 
> Note that I stopped calling ensureResourceStatisticsForPrimaryDomain() here. Instead, I merely construct a ResourceLoadStatistics...

'ensureResourceStatisticsForPrimaryDomain()' only returned a new ResourceLoadStatistics if it didn't already exist.

I think your revised code prevents us from properly counting, because we never retrieve an existing ResourceLoadStatistic object for a domain so that we can update its counts.

Looks like we don't have a good test for that behavior!

[Chris Dumez]

Reverted r218944 for reason:

Optimization is incorrect

Committed r218955: <http://trac.webkit.org/changeset/218955>

[Chris Dumez]

Created attachment 314154 [details]
Patch

[Brent Fulgham]

Comment on attachment 314154 [details]
Patch

r=me

[WebKit Commit Bot]

Comment on attachment 314154 [details]
Patch

Clearing flags on attachment: 314154

Committed r218966: <http://trac.webkit.org/changeset/218966>

[WebKit Commit Bot]

All reviewed patches have been landed.  Closing bug.