1358066 - If the painted layer is optimized to color or image layer, the calculation of invalidate region is wrong.

Status: RESOLVED FIXED

(Core :: Layout, enhancement)

Description

[Morris Tseng [:mtseng] [:Morris] (Inactive)]

I found this while debugging bug 1358055. This would cause infinite loop in reftest. The log showed "timed out waiting for pending paint count to reach zero (after reftest-wait removed and switch to print mode) (waiting for MozAfterPaint)". After debugging it, I found the problem is if the painted layer is optimized to color or image layer, we still calculate the invalidate region by using the corresponding color or image display-item. Which is wrong because the painted layer is not visible and should not produce any invalidate region.

Comment 1

[Morris Tseng [:mtseng] [:Morris] (Inactive)]

Attachment: Bug 1358066 - Remove the painted layer invalidate region calculation if the painted layer has been optimized to color or image layer.

Review commit: https://reviewboard.mozilla.org/r/131994/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/131994/

Comment 2

[Morris Tseng [:mtseng] [:Morris] (Inactive)]

The try result.
https://treeherder.mozilla.org/#/jobs?repo=try&revision=feca2e1a946f70c6adfe09a83b8a78b2a1784be5

Comment 3

[Matt Woodrow (:mattwoodrow)]

Comment on attachment 8859945 [details]
Bug 1358066 - Remove the painted layer invalidate region calculation if the painted layer has been optimized to color or image layer.

https://reviewboard.mozilla.org/r/131994/#review135086

::: layout/painting/FrameLayerBuilder.cpp:1995
(Diff revision 1)
>    }
>  
>    DisplayItemData* data = GetDisplayItemDataForManager(aItem, aLayer->Manager());
>    NS_ASSERTION(data, "Must have already stored data for this item!");
>    data->mOptLayer = aLayer;
> +  data->mItem = nullptr;

Rather than clearing mItem here, can you instead check for the presence of mOptLayer in ComputeGeometryChangeForItem? Please also add a comment there about how we don't need to compute invalidations when we're not using the PaintedLayer.

r=me with that.

Comment 4

[Morris Tseng [:mtseng] [:Morris] (Inactive)]

Comment on attachment 8859945 [details]
Bug 1358066 - Remove the painted layer invalidate region calculation if the painted layer has been optimized to color or image layer.

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/131994/diff/1-2/

Comment 5

[Morris Tseng [:mtseng] [:Morris] (Inactive)]

latest try:
  https://treeherder.mozilla.org/#/jobs?repo=try&revision=09590640e602a8033a0582005d4118ea6a1d66c2

Comment 6

[Morris Tseng [:mtseng] [:Morris] (Inactive)]

Oops, forgot to change reviewer...

Comment 7

[Morris Tseng [:mtseng] [:Morris] (Inactive)]

Comment on attachment 8859945 [details]
Bug 1358066 - Remove the painted layer invalidate region calculation if the painted layer has been optimized to color or image layer.

Review request updated; see interdiff: https://reviewboard.mozilla.org/r/131994/diff/2-3/

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/3a7c9793ce9d
Remove the painted layer invalidate region calculation if the painted layer has been optimized to color or image layer. r=mattwoodrow

Comment 9

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/3a7c9793ce9d