Many composition tests (for example test_2d.composite.solid.color.html) fail under SkiaGL.

What do the failures look like? Are they completely wrong, or off by a bit or two?

(In reply to Matt Woodrow (:mattwoodrow) from comment #1) > What do the failures look like? > > Are they completely wrong, or off by a bit or two? From what I've seen they are off by a decent amount. For instance, on test_2d.composite.solid.lighter.html, it expects 255,255,0,255 but got 0,255,0,255.

Right, I think that rules out it being a rounding bug in the implementation of the blending operators.

Oops that result was for the solid.multiply test. Stupid page titles are wrong. Still, way busted.

and http://mxr.mozilla.org/mozilla-central/source/gfx/skia/src/core/SkXfermode.cpp#639 Looks like the GPU code just fails quietly if it hits a blend mode it doesn't know how to handle. Enjoy :)

(In reply to Matt Woodrow (:mattwoodrow) from comment #5) > http://mxr.mozilla.org/mozilla-central/source/gfx/skia/src/gpu/SkGpuDevice. > cpp#522 I initially saw that too, but that case wasn't being hit when I looked into it. I can look again, though.

Actually, maybe it is in fact hitting that case. Anyway, Chrome fails the multiply one the same way too. We can probably just remove a bunch of the composite tests I guess?

I think all of those advanced operators aren't part of the spec yet (but they are proposed). Adobe might be sad if we stop supporting them again.

The complete set of composite test failures are below. I don't think any of these are in the current spec, are they? George didn't you say someone from Adobe was working on support for this in SkiaGL? 224 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.color-burn.html | pixel 50,25 is 218,255,36,223; expected 108,255,146,223 +/- 5 227 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.color-dodge.html | pixel 50,25 is 218,255,36,223; expected 108,255,146,223 +/- 5 233 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.darken.html | pixel 50,25 is 0,255,255,128; expected 108,255,36,223 +/- 5 242 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.difference.html | pixel 50,25 is 218,255,36,223; expected 218,145,146,223 +/- 5 245 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.exclusion.html | pixel 50,25 is 218,255,36,223; expected 218,145,146,223 +/- 5 251 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.hue.html | pixel 50,25 is 218,255,36,223; expected 194,230,36,223 +/- 5 254 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.lighten.html | pixel 50,25 is 218,255,36,223; expected 218,255,146,223 +/- 5 257 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.luminosity.html | pixel 50,25 is 218,255,36,223; expected 176,255,146,223 +/- 5 260 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.multiply.html | pixel 50,25 is 218,255,36,223; expected 108,255,36,223 +/- 5 263 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.overlay.html | pixel 50,25 is 218,255,36,223; expected 108,255,146,223 +/- 5 266 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.saturation.html | pixel 50,25 is 218,255,36,223; expected 108,253,145,223 +/- 5 269 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.canvas.screen.html | pixel 50,25 is 218,255,36,223; expected 218,255,146,223 +/- 5 293 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.color-burn.html | pixel 50,25 is 255,255,0,255; expected 0,255,255,255 +/- 5 296 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.color-dodge.html | pixel 50,25 is 255,255,0,255; expected 0,255,255,255 +/- 5 299 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.color.html | pixel 50,25 is 255,255,0,255; expected 200,200,0,255 +/- 5 302 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.darken.html | pixel 50,25 is 0,255,255,255; expected 0,255,0,255 +/- 5 305 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.difference.html | pixel 50,25 is 255,255,0,255; expected 255,0,255,255 +/- 5 308 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.exclusion.html | pixel 50,25 is 255,255,0,255; expected 255,0,255,255 +/- 5 314 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.hue.html | pixel 50,25 is 255,255,0,255; expected 200,200,0,255 +/- 5 317 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.lighten.html | pixel 50,25 is 255,255,0,255; expected 255,255,255,255 +/- 5 320 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.luminosity.html | pixel 50,25 is 255,255,0,255; expected 158,255,255,255 +/- 5 323 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.multiply.html | pixel 50,25 is 255,255,0,255; expected 0,255,0,255 +/- 5 326 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.overlay.html | pixel 50,25 is 255,255,0,255; expected 0,255,255,255 +/- 5 329 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.saturation.html | pixel 50,25 is 255,255,0,255; expected 0,254,254,255 +/- 5 332 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.screen.html | pixel 50,25 is 255,255,0,255; expected 255,255,255,255 +/- 5 335 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.composite.solid.soft-light.html | pixel 50,25 is 255,255,0,255; expected 0,255,255,255 +/- 5

If they're not part of the spec yet, we can always mark them as expected failure with SkiaGL, no?

And yeah, Rik Cabanier should probably be CCd on this bug...

(In reply to George Wright (:gw280) from comment #11) > If they're not part of the spec yet, we can always mark them as expected > failure with SkiaGL, no? How old is your version of Skia? The code that adds blending for the GPU was only recently added.

r8495, around a couple of months old

(In reply to George Wright (:gw280) from comment #14) > r8495, around a couple of months old They landed at the end of April. Brian Salomon should know exactly when.

Looks like it landed in r8768 and r8815. I just tried cherry picking the patches over to r8495 but it's non trivial as there's been a lot of code churn upstream. I have a further rebase branch based off r9475 that's mostly working which I did a couple of weeks ago as a test. It may be easier to pull that in than to cherry-pick in the changes that implement these blending modes. I can kick off a try build with it to see what it's like. Brad, what's your opinion?

Here's a try run I just kicked off with Skia r9475 and all the patches currently in the Graphics branch applied on top of it. It should make for some interesting results, especially now that all the memory leaks are fixed. https://tbpl.mozilla.org/?tree=Try&rev=5137a44dc282 Hopefully all the composition failures will be fixed in this rev...

Good news! Whilst the try run was a no-go due to a bad patch I had left in there (and I can't re-push right now as try is closed...), I ran it locally, and got this: 222 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.clip.winding.html | pixel 50,50 is 0,0,0,255; expected 0,255,0,255 +/- 5 223 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.clip.winding.html | pixel 50,50 is 0,0,0,255; expected 255,0,0,255 +/- 5 411 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.front.html | pixel 1,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 412 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.front.html | pixel 50,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 413 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.front.html | pixel 98,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 414 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.front.html | pixel 1,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 415 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.front.html | pixel 50,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 416 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.front.html | pixel 98,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 417 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.front.html | pixel 1,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 418 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.front.html | pixel 50,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 419 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.front.html | pixel 98,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 422 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.top.html | pixel 1,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 423 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.top.html | pixel 50,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 424 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.top.html | pixel 98,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 425 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.top.html | pixel 1,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 426 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.top.html | pixel 50,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 427 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.top.html | pixel 98,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 428 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.top.html | pixel 1,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 429 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.top.html | pixel 50,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 430 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.cone.top.html | pixel 98,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 433 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside2.html | pixel 1,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 434 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside2.html | pixel 50,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 435 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside2.html | pixel 98,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 436 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside2.html | pixel 1,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 437 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside2.html | pixel 50,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 438 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside2.html | pixel 98,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 439 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside2.html | pixel 1,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 440 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside2.html | pixel 50,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 441 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside2.html | pixel 98,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 444 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside3.html | pixel 1,1 is 255,0,0,255; expected 0,255,0,255 +/- 0 445 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside3.html | pixel 50,1 is 255,0,0,255; expected 0,255,0,255 +/- 0 446 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside3.html | pixel 98,1 is 255,0,0,255; expected 0,255,0,255 +/- 0 447 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside3.html | pixel 1,25 is 255,0,0,255; expected 0,255,0,255 +/- 0 448 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside3.html | pixel 50,25 is 255,0,0,255; expected 0,255,0,255 +/- 0 449 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside3.html | pixel 98,25 is 255,0,0,255; expected 0,255,0,255 +/- 0 450 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside3.html | pixel 1,48 is 255,0,0,255; expected 0,255,0,255 +/- 0 451 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside3.html | pixel 50,48 is 255,0,0,255; expected 0,255,0,255 +/- 0 452 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.inside3.html | pixel 98,48 is 255,0,0,255; expected 0,255,0,255 +/- 0 455 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.outside1.html | pixel 1,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 456 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.outside1.html | pixel 50,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 457 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.outside1.html | pixel 98,1 is 1,254,0,255; expected 0,255,0,255 +/- 0 458 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.outside1.html | pixel 1,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 459 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.outside1.html | pixel 50,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 460 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.outside1.html | pixel 98,25 is 1,254,0,255; expected 0,255,0,255 +/- 0 461 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.outside1.html | pixel 1,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 462 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.outside1.html | pixel 50,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 463 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_2d.gradient.radial.outside1.html | pixel 98,48 is 1,254,0,255; expected 0,255,0,255 +/- 0 2987 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 50,25 of c501 is 0,0,255,255; expected 127,0,127,255 +/- 2 2988 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 50,25 of c502 is 75,0,180,255; expected 127,0,127,255 +/- 2 2989 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 50,25 of c503 is 0,0,255,255; expected 127,0,127,255 +/- 2 3022 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 50,25 of c514 is 0,0,255,255; expected 127,0,127,255 +/- 2 3024 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 50,25 of c516 is 255,0,0,255; expected 0,255,0,255 +/- 0 3025 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 25,25 of c517 is 255,0,0,255; expected 0,255,0,255 +/- 0 3036 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 50,25 of c525 is 0,0,255,255; expected 127,0,127,255 +/- 2 3038 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 50,25 of c527 is 255,0,0,255; expected 0,255,0,255 +/- 0 3042 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 50,25 of c529 is 0,0,255,255; expected 127,0,127,255 +/- 2 3047 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 25,25 of c532 is 255,0,0,255; expected 0,255,0,255 +/- 2 3050 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 50,25 of c533 is 255,0,0,255; expected 0,255,0,255 +/- 0 3051 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 25,25 of c534 is 255,0,0,255; expected 0,255,0,255 +/- 0 3066 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 50,25 of c540 is 0,0,255,255; expected 127,0,127,255 +/- 2 3068 ERROR TEST-UNEXPECTED-FAIL | /tests/content/canvas/test/test_canvas.html | pixel 50,25 of c542 is 255,0,0,255; expected 0,255,0,255 +/- 0

Oh dear, not radial gradients again. I assume that means the blend mode tests passed (haven't memorized the test_canvas numbers yet)?

It looks like all the test_canvas failures are either gradials or shadows, so no new issues there.

George tells me on irc that these failures are due to advanced blend ops that aren't standardized yet. Assuming that is true, let's just mark them as expected fail and move on for now.

(In reply to Brad Lassey [:blassey] from comment #21) > George tells me on irc that these failures are due to advanced blend ops > that aren't standardized yet. Assuming that is true, let's just mark them as > expected fail and move on for now. what failures?

(In reply to Rik Cabanier from comment #22) > what failures? They're all listed in comment 10

(In reply to Brad Lassey [:blassey] from comment #21) > George tells me on irc that these failures are due to advanced blend ops > that aren't standardized yet. Assuming that is true, let's just mark them as > expected fail and move on for now. Yes, but we only need to upstream skia to get these working. We pass these tests on all platforms currently (mainly thanks to work from Rik), so I don't think it's fair to break this again. Looks like George has the skia update almost done, we should just finish that.

I think it's more a question of risk. Bringing in the rebase-next to this timeframe brings in an undue amount of risk for the rewards, imho. I think it should be our priority to get what we have working, then aim to ship the latest skia code as soon as possible after shipping the initial GPU-accelerated canvas on Android/Linux/OSX.

(In reply to George Wright (:gw280) from comment #25) > I think it's more a question of risk. Bringing in the rebase-next to this > timeframe brings in an undue amount of risk for the rewards, imho. > > I think it should be our priority to get what we have working, then aim to > ship the latest skia code as soon as possible after shipping the initial > GPU-accelerated canvas on Android/Linux/OSX. if you leave this out you will break pdf.js and your flash emulator

Comment on attachment 766758 [details] [diff] [review] Disable unimplemented canvas composition tests with SkiaGL I don't think this regression is acceptable.

(In reply to Rik Cabanier from comment #26) > (In reply to George Wright (:gw280) from comment #25) > > I think it's more a question of risk. Bringing in the rebase-next to this > > timeframe brings in an undue amount of risk for the rewards, imho. > > > > I think it should be our priority to get what we have working, then aim to > > ship the latest skia code as soon as possible after shipping the initial > > GPU-accelerated canvas on Android/Linux/OSX. > > if you leave this out you will break pdf.js and your flash emulator Understood, but we are only going to enable this for Android and B2G at first. Neither pdf.js or shumway are shipping there right now, so I think we're ok with those being busted for 1.0.

(In reply to :Ms2ger from comment #28) > Comment on attachment 766758 [details] [diff] [review] > Disable unimplemented canvas composition tests with SkiaGL > > I don't think this regression is acceptable. I understand. Someone (not me) is going to have to make a call, I'm just putting the patch up :)

Comment on attachment 766758 [details] [diff] [review] Disable unimplemented canvas composition tests with SkiaGL Review of attachment 766758 [details] [diff] [review]: ----------------------------------------------------------------- ::: content/canvas/test/Makefile.in @@ +119,5 @@ > + test_2d.composite.solid.overlay.html \ > + test_2d.composite.solid.saturation.html \ > + test_2d.composite.solid.screen.html \ > + test_2d.composite.solid.soft-light.html \ > + $(NULL) Do we really want to blanket disable these for all Android/Gonk builds? Is there a way to disable them inside the tests themselves, a little like how test_canvas.html does with IsAzureSkia() in the JS code?

(In reply to George Wright (:gw280) from comment #31) > Comment on attachment 766758 [details] [diff] [review] > Disable unimplemented canvas composition tests with SkiaGL > > Review of attachment 766758 [details] [diff] [review]: > ----------------------------------------------------------------- > > ::: content/canvas/test/Makefile.in > @@ +119,5 @@ > > + test_2d.composite.solid.overlay.html \ > > + test_2d.composite.solid.saturation.html \ > > + test_2d.composite.solid.screen.html \ > > + test_2d.composite.solid.soft-light.html \ > > + $(NULL) > > Do we really want to blanket disable these for all Android/Gonk builds? Is > there a way to disable them inside the tests themselves, a little like how > test_canvas.html does with IsAzureSkia() in the JS code? It would be kind of a PITA I think, but we could do it. Hopefully we don't have these disabled for very long, though, and we can just revert this after we rebase on the newer Skia.

By Murphy's Law, if we disable this on platforms where they're passing, they will start failing. How much PITA would it be? :)

(In reply to Milan Sreckovic [:milan] from comment #33) > By Murphy's Law, if we disable this on platforms where they're passing, they > will start failing. How much PITA would it be? :) We are only disabling them on platforms on which they would fail because SkiaGL is enabled. Any platforms where they currently run will continue to run modulo the SkiaGL ones (Android/Gonk).

(In reply to James Willcox (:snorp) (jwillcox@mozilla.com) from comment #29) > Neither pdf.js or shumway are shipping there right now, so I think > we're ok with those being busted for 1.0. Really? I thought B2G internal PDF Viewer was pdf.js.

(In reply to Masatoshi Kimura [:emk] from comment #35) > (In reply to James Willcox (:snorp) (jwillcox@mozilla.com) from comment #29) > > Neither pdf.js or shumway are shipping there right now, so I think > > we're ok with those being busted for 1.0. > > Really? I thought B2G internal PDF Viewer was pdf.js. Oh, that probably true. In that case we could only turn do Android until we get the rebase. Or just deal with pdf.js being busted on B2G for a few weeks or whatever.

So, when I ran the my version of the graphics branch with skia r9475, I recommended against its use because of a bunch of failing image encoder reftests. This seems to affect r8495 as well, and my try run (https://tbpl.mozilla.org/?tree=Try&rev=7b6bc96f3b1b) of r9475 doesn't show any additional test failures compared to r8495 as a result. I have filed bug 886685 re. the image encoder failures. If we've decided to prioritise the advanced blending modes, it shouldn't really be any additional work now to switch to r9475, given that the delta in terms of test failures is either close to or at 0.

Even in m-c, B2G is sensitive to regressions, as there is a number of people that are developing in it - sounds like we can avoid the regressions with bug 886685, so let's do that.