Closed Bug 590864 Opened 14 years ago Closed 14 years ago

Poor font antialiasing compared to Firefox 3.6.8 in Ubuntu 10.04 LTS

Categories

(Core :: Graphics, defect)

x86_64
Linux
defect
Not set
normal

Tracking

()

RESOLVED DUPLICATE of bug 404637

People

(Reporter: mikko.rantalainen, Unassigned)

References

()

Details

Attachments

(4 files)

User-Agent:       Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.2.8) Gecko/20100723 Ubuntu/10.04 (lucid) Firefox/3.6.8
Build Identifier: Mozilla/5.0 (X11; Linux x86_64; rv:2.0b5pre) Gecko/20100825 Minefield/4.0b5pre

Font antialising results in poor font rendering in trunk builds. The version I tested was http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/latest-trunk/firefox-4.0b5pre.en-US.linux-x86_64.tar.bz2 at 2010-08-26 11:19:25.


Reproducible: Always

Steps to Reproduce:
1. Open ftp://ftp.mozilla.org/index.html

Actual Results:  
The text on the page is rendered with poor quality. See attached screenshots.

Expected Results:  
The text should look equal or better to Firefox 3.6.8.

Correct rendering with Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.2.8) Gecko/20100723 Ubuntu/10.04 (lucid) Firefox/3.6.8.

I'm not sure if the rendering is incorrect in every case or if the Minefield just does not follow system settings for font rendering. The settings I have selected with my Gnome desktop:
System - Preferences - Appearance - Fonts - Details:
Smoothing: Subpixel
Hinting: Medium
Subpixel order: RGB
Possibly related to bug 588407
The Minefield clearly tries to do some kind of subpixel antialiasing but the results are far from great.
Possibly related to bug 363861 or bug 579276? (Both on win32, this one on Linux/x86_64.)
Version: unspecified → Trunk
Could be bug 579276. I'm not sure what the story is with antialiasing on Linux.
It seems that bug 579276 is a bit different: on win32 it causes poor antialiasing if position:absolute is used. This bug is about all text rendered with poor antialiasing (the problems seem similar to rendering of bug 579276 but also position:static looks poor).
I doubt this is related to retained layers or opacity.

The hinting and/or filters look different.

How does a mozilla 3.6.8 build look?
(In reply to comment #7)
> How does a mozilla 3.6.8 build look?

What is "mozilla 3.6.8 build"?

The Firefox 3.6.8 distributed from Ubuntu repositories renders text identical to "native" gnome applications. For example, the label "Help" in the main menu looks identical to label "Help" in Gnome gcalctool (Gnome calculator) or Gnome Character Map. The rendering in trunk minefield does not match and results in poor quality display. See the attached screenshots and bug description for details. Note that the poor rendering affects both content and chrome.
Ubuntu builds typically have Ubuntu customizations, the same customizations as in other Gnome apps on Ubuntu.
Minefield/4.0b5pre would not have these.

There are possibly other factors involved here, but first I want to know whether there is a regression in builds from Mozilla (without Ubuntu tweaks).

Such 3.6.8 builds are available from http://www.mozilla.com/.
Since b5pre, D2D and DW are enabled by default.
You can disable HW acceleration in Options > Advanced > General.

Is the font rendering OK with HW acceleration disabled ?
Could you provide the graphic info in the "about:support" page ?
Component: General → Graphics
Product: Firefox → Core
QA Contact: general → thebes
> Since b5pre, D2D and DW are enabled by default.

How could that possibly affect anything on Ubuntu?
(In reply to comment #9)
> Ubuntu builds typically have Ubuntu customizations, the same customizations as
> in other Gnome apps on Ubuntu.
> Minefield/4.0b5pre would not have these.
> 
> There are possibly other factors involved here, but first I want to know
> whether there is a regression in builds from Mozilla (without Ubuntu tweaks).
> 
> Such 3.6.8 builds are available from http://www.mozilla.com/.

OK. I checked the default version suggested by mozilla.com and I got 32 bit build for linux as firefox-3.6.8.tar.bz2. See the attached screenshot (running on Ubuntu 10.04 LTS x86_64 with 32 bit support libraries). It seems that the rendering looks the same as with 4.0b5pre and the poor rendering quality is caused by not applying the fixes added by Ubuntu.

Conclusion: there is no linux regression between 3.6.8 and 4.0b5pre.

I'd prefer mozilla to adopt the fixes from Ubuntu to honor Gnome appearance settings for font subpixel rendering, but unless this bug gets morphed for that, I think this bug should be closed.
I checked the patches applied to Ubuntu builds and it seems that Ubuntu applies patch "lp512615_cairo_lcd_filter.patch" that fixes bug 404637. As a result, this should probably be marked as duplicate of bug 404637.

See also:
https://bugzilla.mozilla.org/show_bug.cgi?id=404637
http://bugs.freedesktop.org/show_bug.cgi?id=10301
https://bugs.launchpad.net/bugs/512615
I haven't tried to build firefox with this patch but I believe this is the patch that Ubuntu uses to improve font rendering with cairo. I'm not sure about the license of this patch.
Getting the source in Ubuntu: apt-get source firefox, see subdirectory debian/patches after executing that command.
Looking at the patch, perhaps the issue could be missing "(options->hint_style << 12)" in antialias options (https://bugzilla.mozilla.org/attachment.cgi?id=471775&action=diff#mozilla/gfx/cairo/cairo/src/cairo-font-options.c_sec6) ? It could be that without the patch, the hint_style is not correctly initialized to user preferences, either.
Thanks for investigating and reporting back.
Status: UNCONFIRMED → RESOLVED
Closed: 14 years ago
Resolution: --- → DUPLICATE
(In reply to comment #16)
> Looking at the patch, perhaps the issue could be missing "(options->hint_style
> << 12)" in antialias options

I'm guessing that's just changed from "<< 8" to make room for options->lcd_filter.

Bug 404637 would be the place to discuss this, but Mozilla builds will probably have lcd_filter support when upstream cairo does.
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: