[webkit-reviews] review denied: [Bug 54133] [chromium] Replace tiler data structure with something smarter : [Attachment 83403] Patch

[bugzilla-daemon at webkit.org]

Kenneth Russell <kbr at google.com> has denied Adrienne Walker <enne at google.com>'s
request for review:
Bug 54133: [chromium] Replace tiler data structure with something smarter
https://bugs.webkit.org/show_bug.cgi?id=54133

Attachment 83403: Patch
https://bugs.webkit.org/attachment.cgi?id=83403&action=review

------- Additional Comments from Kenneth Russell <kbr at google.com>
View in context: https://bugs.webkit.org/attachment.cgi?id=83403&action=review

Looks good overall. A couple of minor comments/questions and one small must-fix
bug.

> Source/WebCore/platform/graphics/chromium/LayerTilerChromium.cpp:206
>      for (size_t i = 0; i < m_tiles.size(); ++i) {
> -	   if (m_tiles[i])
> -	       m_unusedTiles.append(m_tiles[i].release());
> +	   m_unusedTiles.append(m_tiles[i].release());
>      }

No braces for one-line bodies.

> Source/WebCore/platform/graphics/chromium/LayerTilerChromium.cpp:223
> +    // To avoid an O(n^2) check for tile existence via tileAt, keep a flag
> +    // vector indexed by tile location marking all the tiles we've seen
already.

Wouldn't it be better to just bite the bullet and use a HashMap whose keys are
the (i,j) pairs of ints and whose values are Tile*'s?

> Source/WebCore/platform/graphics/chromium/LayerTilerChromium.cpp:497
> +    m_tiler->m_tiles[m_idx].release();

This leaks memory! You probably meant to call clear(), but you shouldn't need
to do this at all -- the resize() call after the swap() should take care of
destroying the last OwnPtr<Tile> and therefore its referent.