1036400 - Background tab hit region too small in Nightly

Status: RESOLVED FIXED

(Core :: SVG, defect)

Description

[Markus Stange [:mstange]]

Attachment: illustration

In yesterday's Nightly, the mouse hit region for background tabs is too small.
See the attached screenshot.

phlsa tells me that it's bigger in Beta, so this is hopefully not intentional.

Comment 1

[:Gijs (he/him)]

https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=81691a55e60f&tochange=1dc6b294800d

Comment 2

[:Gijs (he/him)]

Looks like bug 1034115. Jonathan, could you have a look?

Comment 3

[:Gijs (he/him)]

Oh. Hmm. Or bug 1031726, because this is clip:rect, not SVG? Maybe? Mats?

Comment 4

[:Gijs (he/him)]

Backing out both of Mats' commits in that window didn't fix this. I tried backing out all the SVG changes (by now there's also bug 988808) but that no longer compiles. Still, I guess that means this is sooner to do with the SVG changes than with the clip:rect changes. :-\

Comment 5

[Jonathan Watt [:jwatt]]

Probably me. I'll take a look.

Comment 6

[Mats Palmgren (inactive)]

Fwiw, I can't reproduce this in Linux64 Nightly 33.0a1 (2014-07-09).

Comment 7

[Jonathan Watt [:jwatt]]

Attachment: patch

I'll work on knocking up a mochitest for this before landing.

Comment 8

[Robert Longson [:longsonr]]

What's the point of the #if 1 change. You can have an r=longsonr for the other bit but not that part.

Comment 9

[Robert Longson [:longsonr]]

Comment on attachment 8453211 [details] [diff] [review]
patch

># HG changeset patch
># Parent 3a090599c39ef7d6ef0c66af45d55cb163c06f4f
># User Jonathan Watt <[e-mail]>
>diff --git a/layout/svg/nsSVGPathGeometryFrame.cpp b/layout/svg/nsSVGPathGeometryFrame.cpp
>--- a/layout/svg/nsSVGPathGeometryFrame.cpp
>+++ b/layout/svg/nsSVGPathGeometryFrame.cpp
>@@ -445,30 +445,34 @@ nsSVGPathGeometryFrame::GetBBoxContribut
>   SVGBBox bbox;
> 
>   if (aToBBoxUserspace.IsSingular()) {
>     // XXX ReportToConsole
>     return bbox;
>   }
> 
>   RefPtr<DrawTarget> tmpDT;
>+#if 1
> #ifdef XP_WIN
>   // Unfortunately D2D backed DrawTarget produces bounds with rounding errors
>   // when whole number results are expected, even in the case of trivial
>   // calculations. To avoid that and meet the expectations of web content we
>   // have to use a CAIRO DrawTarget. The most efficient way to do that is to
>   // wrap the cached cairo_surface_t from ScreenReferenceSurface():
>   nsRefPtr<gfxASurface> refSurf =
>     gfxPlatform::GetPlatform()->ScreenReferenceSurface();
>   tmpDT = gfxPlatform::GetPlatform()->
>     CreateDrawTargetForSurface(refSurf, IntSize(1, 1));
> #else
>   tmpDT = gfxPlatform::GetPlatform()->ScreenReferenceDrawTarget();
> #endif
>   nsRefPtr<gfxContext> tmpCtx = new gfxContext(tmpDT);
>+#else
>+  nsRefPtr<gfxContext> tmpCtx = new gfxContext(gfxPlatform::GetPlatform()->ScreenReferenceSurface());
>+#endif
> 
>   GeneratePath(tmpCtx, aToBBoxUserspace);
>   tmpCtx->IdentityMatrix();
> 
>   // Be careful when replacing the following logic to get the fill and stroke
>   // extents independently (instead of computing the stroke extents from the
>   // path extents). You may think that you can just use the stroke extents if
>   // there is both a fill and a stroke. In reality it's necessary to calculate
>diff --git a/layout/svg/nsSVGUtils.cpp b/layout/svg/nsSVGUtils.cpp
>--- a/layout/svg/nsSVGUtils.cpp
>+++ b/layout/svg/nsSVGUtils.cpp
>@@ -367,16 +367,20 @@ nsSVGUtils::GetOuterSVGFrameAndCoveredRe
> 
> gfxMatrix
> nsSVGUtils::GetCanvasTM(nsIFrame *aFrame, uint32_t aFor,
>                         nsIFrame* aTransformRoot)
> {
>   // XXX yuck, we really need a common interface for GetCanvasTM
> 
>   if (!aFrame->IsFrameOfType(nsIFrame::eSVG)) {
>+    if (aFor == nsISVGChildFrame::FOR_HIT_TESTING &&
>+        NS_SVGDisplayListHitTestingEnabled()) {
>+      return gfxMatrix();
>+    }
>     return nsSVGIntegrationUtils::GetCSSPxToDevPxMatrix(aFrame);
>   }
> 
>   if (!(aFrame->GetStateBits() & NS_FRAME_IS_NONDISPLAY) &&
>       !aTransformRoot) {
>     if (aFor == nsISVGChildFrame::FOR_PAINTING &&
>         NS_SVGDisplayListPaintingEnabled()) {
>       return nsSVGIntegrationUtils::GetCSSPxToDevPxMatrix(aFrame);

Comment 10

[Jonathan Watt [:jwatt]]

(In reply to Robert Longson from comment #8)
> What's the point of the #if 1 change. You can have an r=longsonr for the
> other bit but not that part.

Oops. That has nothing to do with this bug. ;)

Comment 12

[Jonathan Watt [:jwatt]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/8abc07984f2f

Pushed with mochitest.

Comment 13

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/8abc07984f2f