Created attachment 8847099 [details] ff54bug.rar User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:54.0) Gecko/20100101 Firefox/54.0 Build ID: 20170314004020 Steps to reproduce: Test the submitted demo. Actual results: The blurred element is being stretched and position out of place on blur effect. Expected results: The blurred element should preserve it's properties, only the blurred effect can be applied.
Why is it not the expected effect? How to reproduce the expected effect?
The problem is valid only for ff54 in all other browsers work fine without change the x position. The expected effect image can't be change positions only blur effect can be applied.
I see it is different cases in same version with different profile. Normal if: AzureCanvasBackend skia AzureContentBackend skia AzureFallbackCanvasBackend cairo Offset if: AzureCanvasBackend direct2d 1.1 AzureContentBackend direct2d 1.1 AzureFallbackCanvasBackend skia Regression window: https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=3b7d96464366abd6b5a8c45cfec0444510d2f602&tochange=0c3de6069e8efdce7edf84231fcc5a4b09368885
Hmm. I have no idea why this would even be backend-dependent. Lee, do you have any idea?
8 months ago
Tracking 54/55+ to investigate this regression.
YF, do you see this on 55 too, or just on 54?
(In reply to Boris Zbarsky [:bz] (still a bit busy) (if a patch has no decent message, automatic r-) from comment #6) > YF, do you see this on 55 too, or just on 54? 55 too. Version 55.0a1, Build ID 20170314030215.
Created attachment 8847368 [details] [diff] [review] fix DrawTargetD2D1::GetImageForSurface to support uploads in device-space CopySurface is supposed to operate in device-space. Unfortunately, DrawTargetD2D1::GetImageForSurface always reads from the user-space transform (as per https://dxr.mozilla.org/mozilla-central/source/gfx/2d/DrawTargetD2D1.cpp?q=DrawTargetD2D1%3A%3AGetImageForSurface&redirect_type=single#1879). This causes problems when DrawTargetD2D1::CopySurface attempts to clear the transform on the device context (https://dxr.mozilla.org/mozilla-central/source/gfx/2d/DrawTargetD2D1.cpp?q=DrawTargetD2D1%3A%3ACopySurface&redirect_type=single#381). The partial upload inside GetImageForSurface is basically trying to clip the partial upload in user-space, which is wrong. We need to ignore this transform for this to properly clip the uploaded surface. This patch does that, which was simpler that the alternatives.
Created attachment 8847389 [details] [diff] [review] fix DrawTargetD2D1::GetImageForSurface to support uploads in device-space. r=mchang Added a simplified reftest to make sure this doesn't happen in the future.
Pushed by firstname.lastname@example.org: https://hg.mozilla.org/integration/mozilla-inbound/rev/fa6e1037d65d fix DrawTargetD2D1::GetImageForSurface to support uploads in device-space. r=mchang
Comment on attachment 8847389 [details] [diff] [review] fix DrawTargetD2D1::GetImageForSurface to support uploads in device-space. r=mchang Approval Request Comment [Feature/Bug causing the regression]: bug 1334749 [User impact if declined]: Canvas2D putImageData will not be positioned correctly on Windows. [Is this code covered by automated tests?]: yes, added a new test for this [Has the fix been verified in Nightly?]: yes [Needs manual test from QE? If yes, steps to reproduce]: no [List of other uplifts needed for the feature/fix]: just this [Is the change risky?]: no [Why is the change risky/not risky?]: Just a small change to ensure that the draw target's transformation matrix should not be applied in places it shouldn't have been. Fixes user's testcase and has added reftest to verify nothing broken and everything works. [String changes made/needed]: none
Comment on attachment 8847389 [details] [diff] [review] fix DrawTargetD2D1::GetImageForSurface to support uploads in device-space. r=mchang Fix a regression and include a test. Aurora54+.