Closed Bug 1245061 Opened 8 years ago Closed 8 years ago

canvas context2D antialiasing regression (drawImage V.44 - v.45 + generalized > v.46)

Categories

(Core :: Graphics: Canvas2D, defect)

45 Branch
Unspecified
macOS
defect
Not set
major

Tracking

()

RESOLVED WORKSFORME

People

(Reporter: tristan.fraipont, Unassigned)

Details

(Keywords: regression, Whiteboard: [gfx-noted])

Attachments

(4 files, 1 obsolete file)

Attached file JaggedBorder.html
User Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.9; rv:45.0) Gecko/20100101 Firefox/45.0
Build ID: 20160127070712

Steps to reproduce:

In v.44 to v.45, rotate the context by 45 degrees, then use the drawImage method.
In v.46 rotate the context by 45 degrees, and draw a line or a rect.

online fiddle : https://jsfiddle.net/ty1rnc4x/

[confirmed on mac, I didn't tested on other OS]


Actual results:

The drawn shape has jagged borders


Expected results:

The borders should be smoother.
Severity: normal → major
Component: Untriaged → Canvas: 2D
OS: Unspecified → Mac OS X
Product: Firefox → Core
Can you find a regression window using mozregression?
Flags: needinfo?(tristan.fraipont)
I haven't been able to reproduce a rendering difference on a 2013 MBP w/ OSX 10.11.
Attached image 10.10.3 FF38.tiff (obsolete) —
Screenshot of the expected result as it is in FF38
Flags: needinfo?(tristan.fraipont)
Attached image 10.10.3 FF44.png
Screenshot of the bugged output in FF44
Attachment #8716746 - Attachment mime type: image/png → image/tiff
(In reply to Jeff Muizelaar [:jrmuizel] from comment #1)
> Can you find a regression window using mozregression?

Sorry, I spent the weekend trying to install mozregression-gui without success...


(In reply to Ryan VanderMeulen [:RyanVM] from comment #2)
> I haven't been able to reproduce a rendering difference on a 2013 MBP w/ OSX
> 10.11.

I am able to reproduce it on the same macbook pro Early 2011 (gc  Intel HD Graphics 3000 512 MB) on 10.10 and 10.11 so this is not OS version dependent.
Attachment #8716746 - Attachment is obsolete: true
Attached image 10.10.3 FF38.png
Screenshot of the expected result in FF38
(In reply to Kaiido from comment #5)
> (In reply to Jeff Muizelaar [:jrmuizel] from comment #1)
> > Can you find a regression window using mozregression?
> 
> Sorry, I spent the weekend trying to install mozregression-gui without
> success...

What problem did you run into with mozregression-gui?
Flags: needinfo?(tristan.fraipont)
The testcase looks the same to me comparing Chrome and Firefox 48.0a1.
Kaiido, can you please confirm that you still see this in the latest Firefox Nightly build?
Flags: needinfo?(tristan.fraipont)
Whiteboard: [gfx-noted]
Indeed, it's fixed.
Now drawImage in nightly does produce the exact same pixels as fillRect in v.44 !
fillRect in Nightly does produce an other result but it's not jagged anymore.

For moz-regression, the error I have is "ERROR: No module named sip", I still can't use it, but it's not related with this bug report.

Ps: I don't know how to mark as "Fixed"
Flags: needinfo?(tristan.fraipont)
Thank you for checking. I'm closing this bug as WORKSFORME since I don't see a fix that we made directly to this issue. Please reopen the bug report if the issue returns.
Status: UNCONFIRMED → RESOLVED
Closed: 8 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: