1450262 - SVG feColorMatrix bug with alpha column

Status: UNCONFIRMED

(Core :: Graphics, defect, P3)

Description

[firzen]

Attachment: ex1.svg

User Agent: Mozilla/5.0 (Windows NT 6.1; Win64; x64; rv:59.0) Gecko/20100101 Firefox/59.0
Build ID: 20180323154952

Steps to reproduce:

I was trying to convert the alpha channel of a SVG turbulence to a grayscale (alpha 0 -> black; alpha 1 -> white). Therefore I used a turbulence filter as input to a colorMatrix filter that maps r = 1a, g = 1a, b = 1a, a = 1.

This is fltr2 in the attached ex1.svg. fltr1 is the raw turbulence. I've added some variants for testing purposes as further filters. fltr3 inverse-maps alpha to cyan (alpha 0 -> white; alpha 1 -> red). fltr4 maps alpha to red (alpha 0 -> black; alpha 1 -> red). fltr5 inverse-maps alpha to white (alpha 0 -> white; alpha 1 -> black). fltr6 maps green to white (green 0 -> black; green 1 -> white).

Maybe you can also view it on Gist:
https://gist.github.com/Neonit/d6f8275e426afc978920140573f9830c
(here is the raw link; Gist had problems displaying it to me when I tried: https://gist.githubusercontent.com/Neonit/d6f8275e426afc978920140573f9830c/raw/ee5d1216445a8b0e536d0daa9ea7a6f85d89e81e/ex1.svg)


Actual results:

Firefox 59 (Windows 7 64 Bit) is unable to display fltr2 and fltr4 correctly. Instead, it shows just black.

https://imgur.com/heH3yB7


Expected results:

Chrome 65 and Inkscape 0.92 display it correctly.

https://imgur.com/LmUlN2n

Comment 1

[Robert Longson [:longsonr]]

I think that's a duplicate of bug 1421537

Comment 2

[firzen]

(In reply to Robert Longson [:longsonr] from comment #1)
> I think that's a duplicate of bug 1421537

Just had a look at the referenced issue. Bug 1421537 does not appear for me. Both gooey shapes have a solid color (blue and black). Same machine and browser used for my bug.

Comment 3

[Lee Salzman [:lsalzman]]

firzen, can you try using mozregression to help us figure out which change we might have made that caused this issue?

https://mozilla.github.io/mozregression/quickstart.html

Comment 4

[firzen]

I ran mozregression and got lucky (had to guess a version where it worked last). However, in the end it just output some DEBUGs and WARNINGS (skipping build, unable to find build info using the taskcluster route ...) for some minutes. I guess I could have shrinked the window even further, if these errors wouldn't have occurred.

You can find the full log here: https://gist.github.com/Neonit/18bb2cf908b5837f536011e74067fb98

The relevant line would be this I guess (version 51.0a1):

INFO : Narrowed nightly regression window from [2016-08-08, 2016-08-10] (2 days) to [2016-08-09, 2016-08-10] (1 days) (~0 steps left)

The pushlog_url: https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=720b5d2c84d5b253d4dfde4897e13384dc97a46a&tochange=6cf0089510fad8deb866136f5b92bbced9498447

After that I got a lot of messages in this style:

2018-04-02T10:36:13: DEBUG : Using url: https://hg.mozilla.org/mozilla-central/json-pushes?changeset=720b5d2c84d5b253d4dfde4897e13384dc97a46a
2018-04-02T10:36:14: DEBUG : Using url: https://hg.mozilla.org/mozilla-central/json-pushes?fromchange=720b5d2c84d5b253d4dfde4897e13384dc97a46a&tochange=6cf0089510fad8deb866136f5b92bbced9498447
2018-04-02T10:36:14: DEBUG : using taskcluster route 'gecko.v2.mozilla-central.revision.720b5d2c84d5b253d4dfde4897e13384dc97a46a.firefox.win64-opt'
2018-04-02T10:36:14: DEBUG : using taskcluster route 'gecko.v2.mozilla-central.revision.6cf0089510fad8deb866136f5b92bbced9498447.firefox.win64-opt'
2018-04-02T10:36:22: WARNING : Skipping build 6cf0089510fa: Unable to find build info using the taskcluster route 'gecko.v2.mozilla-central.revision.6cf0089510fad8deb866136f5b92bbced9498447.firefox.win64-opt'
2018-04-02T10:36:22: WARNING : Skipping build 720b5d2c84d5: Unable to find build info using the taskcluster route 'gecko.v2.mozilla-central.revision.720b5d2c84d5b253d4dfde4897e13384dc97a46a.firefox.win64-opt'
2018-04-02T10:36:22: DEBUG : Using url: https://hg.mozilla.org/mozilla-central/json-pushes?startID=30523&endID=30543
2018-04-02T10:36:23: DEBUG : using taskcluster route 'gecko.v2.mozilla-central.revision.763fe887c37cee5fcfe0f00e94fdffc84a41ea1c.firefox.win64-opt'
2018-04-02T10:36:23: DEBUG : using taskcluster route 'gecko.v2.mozilla-central.revision.61a4dd746681fe7c56e1802fc80fe58caf426869.firefox.win64-opt'
2018-04-02T10:36:23: DEBUG : using taskcluster route 'gecko.v2.mozilla-central.revision.e78975b53563d80c99ebfbdf8a9fbf6b829a8a48.firefox.win64-opt'
2018-04-02T10:36:29: WARNING : Skipping build 61a4dd746681: Unable to find build info using the taskcluster route 'gecko.v2.mozilla-central.revision.61a4dd746681fe7c56e1802fc80fe58caf426869.firefox.win64-opt'
2018-04-02T10:36:30: WARNING : Skipping build 763fe887c37c: Unable to find build info using the taskcluster route 'gecko.v2.mozilla-central.revision.763fe887c37cee5fcfe0f00e94fdffc84a41ea1c.firefox.win64-opt'
2018-04-02T10:36:30: WARNING : Skipping build e78975b53563: Unable to find build info using the taskcluster route 'gecko.v2.mozilla-central.revision.e78975b53563d80c99ebfbdf8a9fbf6b829a8a48.firefox.win64-opt'
(AND MORE)

Another maybe useful note might be that I got info messages like the following when opening my example SVG with one of the launched builds where everything was displayed correctly.

2018-04-02T10:36:10: INFO : ?????????????????????: F?????????????????????: F?????????????????????: F[GFX1-]: [D3D11] create component blender failed: 0x80070057, 0
2018-04-02T10:36:12: INFO : ?????????????????????: F?????????????????????: F?????????????????????: F?????????????????????: F?????????????????????: F?????????????????????: F[GFX1-]: [D3D11] create component blender failed: 0x80070057, 0

Comment 5

[torokati44]

The rendered image in the linked gist now matches the "correct" screenshot for me.

System info:
Name - Firefox
Version - 83.0
Build ID - 20201201222123
Distribution ID - fedora
User Agent - Mozilla/5.0 (X11; Fedora; Linux x86_64; rv:83.0) Gecko/20100101 Firefox/83.0
OS - Linux 5.9.11-200.fc33.x86_64 #1 SMP Tue Nov 24 18:18:01 UTC 2020