1366097 - Get vector images working with WebRender

Status: RESOLVED FIXED

(Core :: Graphics: WebRender, enhancement, P1)

Description

[Jeff Muizelaar [:jrmuizel]]

This currently puts the bullets on a wikipedia page in PaintedLayers

Comment 1

[Jeff Muizelaar [:jrmuizel]]

Here's the part that fails:

       nsCOMPtr<imgIContainer> srcImage;
       requestProxy->GetImage(getter_AddRefs(srcImage));
       if (srcImage && !srcImage->IsImageContainerAvailable(aManager, imgIContainer::FLAG_NONE)) {

Comment 2

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v1

Comment 3

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 2. Split out helper functions in VectorImage to get its size and surface., v1

Comment 4

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 3. Move ImageContainer details from RasterImage to ImageResource., v1

Comment 5

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Implement VectorImage::GetImageContainer and VectorImage::IsImageContainerAvailable., v1

Comment 6

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 5. Mark VectorImage as surface as shared when used outside an image layer., v1

try: https://treeherder.mozilla.org/#/jobs?repo=try&revision=069a922c50d79fe0578d8ea66431d5fdf4add15c

Comment 7

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v2

Fix an accidental condition change, should solve the crashes on the previous try.

try: https://treeherder.mozilla.org/#/jobs?repo=try&revision=93944639bb2c30c59487d2d1f826c6e950c75cc5

Comment 8

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Implement VectorImage::GetImageContainer and VectorImage::IsImageContainerAvailable., v2

Add maximum texture size checks to mirror RasterImage.

Comment 9

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 2. Split out helper functions in VectorImage to get its size and surface., v2

Minor fix, shouldn't have an impact, but if somehow GetFrameInternal was called where mFullyLoaded was set, it would incorrectly return BAD_IMAGE instead of NOT_READY. The callers checked this already though.

Comment 10

[Andrew Osmond [:aosmond] (he/him)]

Comment on attachment 8872612 [details] [diff] [review]
Part 4. Implement VectorImage::GetImageContainer and VectorImage::IsImageContainerAvailable., v2

This will be changing as part of bug 1368776 so maybe the review for these bits as not necessary.

Comment 11

[Timothy Nikkel (:tnikkel)]

Comment on attachment 8872591 [details] [diff] [review]
Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v2

>+VectorImage::DrawInternal(const SVGDrawingParameters& aParams,

>   // Set context paint (if specified) on the document:
>   Maybe<AutoSetRestoreSVGContextPaint> autoContextPaint;
>-  if (haveContextPaint && !blockContextPaint) {
>-    autoContextPaint.emplace(aSVGContext->GetContextPaint(),
>+  if (aContextPaint) {
>+    autoContextPaint.emplace(aParams.svgContext->GetContextPaint(),
>                              mSVGDocumentWrapper->GetDocument());

This actually changes us from using aSVGContext always to the new svg context if it exists.

Comment 12

[Andrew Osmond [:aosmond] (he/him)]

(In reply to Timothy Nikkel (:tnikkel) from comment #11)
> Comment on attachment 8872591 [details] [diff] [review]
> Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using
> cached surface., v2
> 
> >+VectorImage::DrawInternal(const SVGDrawingParameters& aParams,
> 
> >   // Set context paint (if specified) on the document:
> >   Maybe<AutoSetRestoreSVGContextPaint> autoContextPaint;
> >-  if (haveContextPaint && !blockContextPaint) {
> >-    autoContextPaint.emplace(aSVGContext->GetContextPaint(),
> >+  if (aContextPaint) {
> >+    autoContextPaint.emplace(aParams.svgContext->GetContextPaint(),
> >                              mSVGDocumentWrapper->GetDocument());
> 
> This actually changes us from using aSVGContext always to the new svg
> context if it exists.

Noted thanks. I already fixed it in the rewrite which should also make it much easier to review :).

Comment 13

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 1. Move RasterImage's ImageContainer state to ImageResource., v3

Comment 14

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 2. Move RasterImage::GetImageContainer and UpdateImageContainer implementations to ImageResource., v3

Comment 15

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 3. Move RasterImage::GetCurrentImage to ImageResource., v3

Comment 16

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Handle all potential DrawResult values to make Image::GetImageContainerImpl more generic., v3

Comment 17

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 5. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v3

Comment 18

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 6. Move RasterImage::mHasSize and mSize to ImageResource., v3

Comment 19

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 7. Make VectorImage use ImageResource::mHasSize and mSize to cache its size information., v3

Comment 20

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 8. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers., v3

Comment 21

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 9. Implement VectorImage::GetImageContainer and VectorImage::IsImageContainerAvailable., v3

Comment 22

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 10. Mark VectorImage as surface as shared when used outside an image layer., v3

Comment 23

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 3. Move RasterImage::GetCurrentImage to ImageResource., v4

Merge error.

Comment 24

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Handle all potential DrawResult values to make Image::GetImageContainerImpl more generic., v4

Comment 25

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 5. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v4

Comment 26

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 6. Move RasterImage::mHasSize and mSize to ImageResource., v4

Merge error.

Comment 27

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 7. Make VectorImage use ImageResource::mHasSize and mSize to cache its size information., v4

Comment 28

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 8. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers., v4

Comment 29

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 9. Implement VectorImage::GetImageContainer and VectorImage::IsImageContainerAvailable., v4

Comment 30

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 10. Mark VectorImage as surface as shared when used outside an image layer., v4

try: https://treeherder.mozilla.org/#/jobs?repo=try&revision=7038bf7ce454143b28df67bd4626225afebaac1a

Comment 31

[Andrew Osmond [:aosmond] (he/him)]

(In reply to Andrew Osmond [:aosmond] from comment #30)
> Created attachment 8874563 [details] [diff] [review]
> Part 10. Mark VectorImage as surface as shared when used outside an image
> layer., v4
> 
> try:
> https://treeherder.mozilla.org/#/
> jobs?repo=try&revision=7038bf7ce454143b28df67bd4626225afebaac1a

It appears that downscale-on-decode is necessary for vector image image container support. Not for downscaling, but for *upscaling*. Without it, images look blurry in the reftest. Sigh.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=f57a27d8e313e0ad557105b85912731bfe2b75ea

Updated part 5 from downscale-on-decode, which forces imgIContainer::GetImageContainer uses from nsDisplayImage and related to use said functionality, fixes at least some of these reftests, or all of them if we are lucky.

Comment 32

[Andrew Osmond [:aosmond] (he/him)]

Okay try again, my branch didn't have the latest mImageContainer changes merged in, but when I pushed to try I updated and broke the build.

try: https://treeherder.mozilla.org/#/jobs?repo=try&revision=1ed8335bbcf61c316af3fdec6099c68b51a47430

Comment 33

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v5

Comment 34

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 2. Move RasterImage::mHasSize and mSize to ImageResource., v5

Comment 35

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 3. Make VectorImage use ImageResource::mHasSize and mSize to cache its size information., v5

Comment 36

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Implement VectorImage::GetFrameInternal as required by Ima geResource for image containers., v5

Comment 37

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 5. Implement VectorImage::GetImageContainer and VectorImage:: IsImageContainerAvailable., v5

Comment 38

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 6. Mark VectorImage as surface as shared when used outside an image layer., v5

Comment 39

[Andrew Osmond [:aosmond] (he/him)]

This requires support for multi-sized image containers being added in bug 1368776. This is particularly important for vector images because not only can they be downscaled, they can also be upscaled, and without that, they may look blurry (because they were rendered at their "native" size, but then the graphics code later had to upscale the rasterized version for the actual display).

Comment 40

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v6

Comment 41

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 2. Move RasterImage::mHasSize and mSize to ImageResource., v6

Comment 42

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 3. Make VectorImage use ImageResource::mHasSize and mSize to cache its size information., v6

Comment 43

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers., v6

Comment 44

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 5. Implement VectorImage::GetImageContainer and VectorImage:: IsImageContainerAvailable., v6

Comment 45

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 6. Mark VectorImage as surface as shared when used outside an image layer., v6

try (without layout changes, non-WR unaffected): https://treeherder.mozilla.org/#/jobs?repo=try&revision=8c37158822d51a261f32b0e8eb4901512912c9a6&selectedJob=113635240

try (with layout changes to use new APIs, fixes WR failures): https://treeherder.mozilla.org/#/jobs?repo=try&revision=247205171456baa9774df2322494b70616fdc41c

Comment 46

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v7

Comment 47

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 2. Move RasterImage::mHasSize and mSize to ImageResource., v7

Comment 48

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 3. Make VectorImage use ImageResource::mHasSize and mSize to cache its size information., v7

Comment 49

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers., v7

Comment 50

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 5. Implement VectorImage::GetImageContainer and VectorImage:: IsImageContainerAvailable., v7

Comment 51

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 6. Mark VectorImage as surface as shared when used outside an image layer., v7

try: https://treeherder.mozilla.org/#/jobs?repo=try&revision=2bc276a3caedb05775b65bfa28f9b02e8fd9ca92

Comment 52

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v8

Comment 53

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 2. Move RasterImage::mHasSize and mSize to ImageResource., v8

Comment 54

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 3. Make VectorImage use ImageResource::mHasSize and mSize to cache its size information., v8

Comment 55

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers., v8

Comment 56

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 5. Implement VectorImage::GetImageContainer and VectorImage:: IsImageContainerAvailable., v8

Comment 57

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 6. Mark VectorImage as surface as shared when used outside an image layer., v8

Comment 58

[Timothy Nikkel (:tnikkel)]

Comment on attachment 8926534 [details] [diff] [review]
Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v8

>@@ -746,33 +746,46 @@ VectorImage::GetFrameAtSize(const IntSize& aSize,
>+  SVGDrawingParameters params(context, aSize, ImageRegion::Create(aSize),
>+                              SamplingFilter::POINT, Nothing(),
>+                              mSVGDocumentWrapper->GetCurrentTime(),
>+                              aFlags, 1.0);

Draw uses 0.0f as the time if the first frame is requested, shouldn't we also?

The fact that we pass in Nothing for the svg context smells like a bug here. Draw uses the svg context to determine if we should use context painting. The fact that we don't have an svg context here suggests that if we layerize vector images with context painting we will break them.

Comment 59

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers., v9

This had a very subtle bug where the SVGDrawingParameters used in GetFrameInternal used "Nothing()" directly for its constructor / Maybe<SVGImageContext> parameter. The problem being that SVGDrawingParameters stores it as a reference instead of making a copy. On debug builds no problem, on optimized, crash. Part 7 will change this again.

Comment 60

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 7. Add an SVGImageContext parameter to imgIContainer::GetImageContainerAtSize., v1

SVGImageContext is sort of important for SVG. This refactors GetImageContainerAtSize to add said parameter so that callers can pass it in to get the container they really expect to get.

Comment 61

[Timothy Nikkel (:tnikkel)]

Comment on attachment 8928658 [details] [diff] [review]
Part 7. Add an SVGImageContext parameter to imgIContainer::GetImageContainerAtSize., v1

Don't we need the svg context for all ways that we can't get image data then? GetFrame, GetFrameAtSize, and GetImageContainer?

Comment 62

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers., v10

Comment 63

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 5. Implement VectorImage::GetImageContainer and VectorImage:: IsImageContainerAvailable., v9

It shouldn't give a container for animated SVGs. Yet anyways.

Comment 64

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 6. Mark VectorImage as surface as shared when used outside an image layer., v9

Make this work as intended -- it was marking surfaces as shared on the image container code path (doh!). Also it wasn't always returning the cached surface (the one that could be a SourceSurfaceSharedData) and instead was using DrawTarget::Snapshot(). There will need to be *another* change to make imgFrame::InitWithDrawable to always use a shared surface (as of right now it only does so on Android and Mac OS X due to the gfxPlatform::CanRenderContentToDataSurface call).

Comment 65

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 7. Add an SVGImageContext parameter to imgIContainer::GetImageContainerAtSize., v2

Rebased.

(In reply to Timothy Nikkel (:tnikkel) from comment #61)
> Comment on attachment 8928658 [details] [diff] [review]
> Part 7. Add an SVGImageContext parameter to
> imgIContainer::GetImageContainerAtSize., v1
> 
> Don't we need the svg context for all ways that we can't get image data
> then? GetFrame, GetFrameAtSize, and GetImageContainer?

For the moment, I'm happy to deprecate those APIs. The WebRender path only requires GetImageContainerAtSize as far as I know, at this very moment. To see how we are using these new APIs, check out the patches in bug 1183378.

Comment 66

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 8. Restrict use of SVGImageContext in VectorImage::GetImageContainerAtSize to permitted URIs., v1

The SVGImageContext is actually restricted, so we need to abstract out the checks from Draw into its own function that can be shared with GetFrameInternal.

Comment 67

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 9. VectorImage::Show should delegate to ImageResource::NotifyDrawingObservers., v1

In the RasterImage image container refactoring, we already moved the NotifyDrawingObservers API to a shared location. Let's actually use that here.

Comment 68

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 8. Restrict use of SVGImageContext in VectorImage::GetImageContainerAtSize to permitted URIs., v2

Update this so that it masks off FLAG_FORCE_PRESERVEASPECTRATIO_NONE before passing to GetImageContainerImpl. This is important because it will still match an image container with a new SVG context after applying the flag as another call with the same SVG context (after application) without the flag.

Comment 69

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 10. Improve image memory reports to include SVG context information., v1

I was thinking this might help debug the remaining failures but thus far, no dice. Making the SVG context info match still results in different results from VectorImage::GetImageContainerAtSize vs Draw. Still useful stuff for the future probably.

Comment 70

[Andrew Osmond [:aosmond] (he/him)]

(In reply to Timothy Nikkel (:tnikkel) from comment #11)
> Comment on attachment 8872591 [details] [diff] [review]
> Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using
> cached surface., v2
> 
> >+VectorImage::DrawInternal(const SVGDrawingParameters& aParams,
> 
> >   // Set context paint (if specified) on the document:
> >   Maybe<AutoSetRestoreSVGContextPaint> autoContextPaint;
> >-  if (haveContextPaint && !blockContextPaint) {
> >-    autoContextPaint.emplace(aSVGContext->GetContextPaint(),
> >+  if (aContextPaint) {
> >+    autoContextPaint.emplace(aParams.svgContext->GetContextPaint(),
> >                              mSVGDocumentWrapper->GetDocument());
> 
> This actually changes us from using aSVGContext always to the new svg
> context if it exists.

Hm. But the new SVG context is a copy of the old one with the ContextPaint filtered out if blockContextPaint == true. It shouldn't have a behaviour change.

Comment 71

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v9

Fix how GetFrameAtSize should pass Nothing() to the SVGDrawingParameters, as it keeps a reference, not a copy. Instead put SVGImageContext on the stack and take a reference to *that*.

Comment 72

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface., v10

(In reply to Timothy Nikkel (:tnikkel) from comment #58)
> Comment on attachment 8926534 [details] [diff] [review]
> Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using
> cached surface., v8
> 
> >@@ -746,33 +746,46 @@ VectorImage::GetFrameAtSize(const IntSize& aSize,
> >+  SVGDrawingParameters params(context, aSize, ImageRegion::Create(aSize),
> >+                              SamplingFilter::POINT, Nothing(),
> >+                              mSVGDocumentWrapper->GetCurrentTime(),
> >+                              aFlags, 1.0);
> 
> Draw uses 0.0f as the time if the first frame is requested, shouldn't we
> also?
> 

Fixed.

> The fact that we pass in Nothing for the svg context smells like a bug here.
> Draw uses the svg context to determine if we should use context painting.
> The fact that we don't have an svg context here suggests that if we layerize
> vector images with context painting we will break them.

You are correct. Parts 7 and 8 should address this.

Comment 73

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers., v11

Rebase.

Comment 74

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers., v12

Too many patches in my queue. Sigh. *Real* patch this time.

Comment 75

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 6. Mark VectorImage as surface as shared when used outside an image layer., v10

Comment 76

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 7. Add an SVGImageContext parameter to imgIContainer::GetImageContainerAtSize., v3

Rebase.

Comment 77

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 8. Restrict use of SVGImageContext in VectorImage::GetImageContainerAtSize to permitted URIs., v3

Comment 78

[Timothy Nikkel (:tnikkel)]

Comment on attachment 8926536 [details] [diff] [review]
Part 3. Make VectorImage use ImageResource::mHasSize and mSize to cache its size information., v8

Talked to Andrews, the plan is to skip this patch as it doesn't seem important.

Comment 79

[Timothy Nikkel (:tnikkel)]

Comment on attachment 8929250 [details] [diff] [review]
Part 4. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers., v12

Can we keep the RecoverFromLostSurfaces call in LookupCacheSurface so all callers don't have to repeat it?

Comment 80

[Timothy Nikkel (:tnikkel)]

Comment on attachment 8929251 [details] [diff] [review]
Part 6. Mark VectorImage as surface as shared when used outside an image layer., v10

It's quite confusing to call MarkSurfaceShared to say that the surface isn't going to be shared.

Comment 81

[Timothy Nikkel (:tnikkel)]

(In reply to Andrew Osmond [:aosmond] from comment #65)
> (In reply to Timothy Nikkel (:tnikkel) from comment #61)
> > Comment on attachment 8928658 [details] [diff] [review]
> > Part 7. Add an SVGImageContext parameter to
> > imgIContainer::GetImageContainerAtSize., v1
> > 
> > Don't we need the svg context for all ways that we can't get image data
> > then? GetFrame, GetFrameAtSize, and GetImageContainer?
> 
> For the moment, I'm happy to deprecate those APIs. The WebRender path only
> requires GetImageContainerAtSize as far as I know, at this very moment. To
> see how we are using these new APIs, check out the patches in bug 1183378.

Okay, so either we need a bug filed to either convert or deprecate these functions.

Comment 82

[Andrew Osmond [:aosmond] (he/him)]

Attachment: Part 4. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers., v13 [carries r=tnikkel]

(In reply to Timothy Nikkel (:tnikkel) from comment #79)
> Comment on attachment 8929250 [details] [diff] [review]
> Part 4. Implement VectorImage::GetFrameInternal as required by ImageResource
> for image containers., v12
> 
> Can we keep the RecoverFromLostSurfaces call in LookupCacheSurface so all
> callers don't have to repeat it?

I can do one better. I'm sure changed the prototype of LookupCacheSurface made sense at one point, but now I've reverted it back because it added no value as this currently stand.

Comment 83

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/eb77f75f95d9
Part 1. VectorImage::GetFrameAtSize should not create a DrawTarget if using cached surface. r=tnikkel
https://hg.mozilla.org/integration/mozilla-inbound/rev/12fc7f8b8930
Part 2. Implement VectorImage::GetFrameInternal as required by ImageResource for image containers. r=tnikkel
https://hg.mozilla.org/integration/mozilla-inbound/rev/a4b811e0b860
Part 3. Implement VectorImage::GetImageContainerAtSize and VectorImage::IsImageContainerAvailableAtSize. r=tnikkel
https://hg.mozilla.org/integration/mozilla-inbound/rev/79b165219828
Part 4. Mark VectorImage as surface as shared when used outside an image layer. r=tnikkel
https://hg.mozilla.org/integration/mozilla-inbound/rev/54d6ad659f11
Part 5. Add an SVGImageContext parameter to imgIContainer::GetImageContainerAtSize. r=tnikkel
https://hg.mozilla.org/integration/mozilla-inbound/rev/e65c223336d9
Part 6. Restrict use of SVGImageContext in VectorImage::GetImageContainerAtSize to permitted URIs. r=tnikkel
https://hg.mozilla.org/integration/mozilla-inbound/rev/71c36630c141
Part 7. VectorImage::Show should delegate to ImageResource::NotifyDrawingObservers. r=tnikkel
https://hg.mozilla.org/integration/mozilla-inbound/rev/3395c151dabb
Part 8. Improve image memory reports to include SVG context information. r=tnikkel

Comment 84

[Noemi Erli[:noemi_erli]]

https://hg.mozilla.org/mozilla-central/rev/eb77f75f95d9
https://hg.mozilla.org/mozilla-central/rev/12fc7f8b8930
https://hg.mozilla.org/mozilla-central/rev/a4b811e0b860
https://hg.mozilla.org/mozilla-central/rev/79b165219828
https://hg.mozilla.org/mozilla-central/rev/54d6ad659f11
https://hg.mozilla.org/mozilla-central/rev/e65c223336d9
https://hg.mozilla.org/mozilla-central/rev/71c36630c141
https://hg.mozilla.org/mozilla-central/rev/3395c151dabb