FS#2033 - freetype bytecode patch not needed

Attached to Project: Arch Linux
Opened by Aaron Griffin (phrakture) - Tuesday, 18 January 2005, 23:01 GMT
Last edited by arjan timmerman (blaasvis) - Sunday, 06 March 2005, 19:15 GMT
Task Type Bug Report
Category System
Status Closed
Assigned To Judd Vinet (judd)
Architecture not specified
Severity High
Priority Normal
Reported Version 0.7 Wombat
Due in Version Undecided
Due Date Undecided
Percent Complete 0%
Votes 0
Private No

Details

The bytecode patch to freetype 2.1.9 (and any freetype for that matter) is flat out NOT needed. This patch specifically has caused firefox to hang and crash numerous times for me (and others). freetype is a very good, very mature piece of software and the devs have stated numerous times that they recommend using the auto-hinter and not the bytecode interpreter...

see here, near the bottom:
http://wiki2.archlinux.org/index.php/XOrg%20Font%20Configuration

I strongly suggest removing the bytecode patch and never using it again - it causes instability and goes against what the developers intend/suggest.
This task depends upon

Closed by  Judd Vinet (judd)
Sunday, 03 April 2005, 19:04 GMT
Reason for closing:  Fixed
Comment by Jan de Groot (JGC) - Wednesday, 19 January 2005, 08:48 GMT
Also, I like the fonts better without the patch. It wouldn't be a problem to remove the patch for me.

Firefox hanging should be solved with latest releases: those are patched for the removal of freetype2 caching code that isn't in freetype2 anymore since 2.1.8
Comment by Aaron Griffin (phrakture) - Wednesday, 19 January 2005, 15:31 GMT
I'll second that. The fonts look much more like Microsoft's ClearType, which looks great on LCD screens. The auto-hinted freetype fonts look like ClearType on all screens, not just LCD.
This would also probably get rid of the "My fonts look ugly" complaints...
Comment by Judd Vinet (judd) - Wednesday, 19 January 2005, 18:16 GMT
Hmmm. on my display, the bytecode interpreter is still needed. I've attached two pictures to show the difference. The autohinter makes my fonts quite a bit more chunky.

Unless it's just a configuration issue on my end.
Comment by Judd Vinet (judd) - Wednesday, 19 January 2005, 18:24 GMT
Whoops, I found my config issue. I still had AA disabled for fonts <= 12 pt.

A new image is up, which looks a lot better.

To be honest, I still like the bytecode version best, as the fonts are crisper and don't hurt my eyes. Staring at the fuzzier AA/auto-hinted ones make my eyes a little sore after a while.

Perhaps the best solution for both worlds (for now) would be to have fonts.conf enable the auto-hinter by default, and then picky people like myself could disable it and revert to the bytecode interpreter if they so choose.
Comment by Aaron Griffin (phrakture) - Wednesday, 19 January 2005, 18:26 GMT
That looks like a problem with the font itself. Is that a MS font? If you can, try with the bitstream vera fonts. I cannot recall the exacts, but I think the MS fonts don't behave well with autohinting. I used to have some writings by Keith Packard about this sort of thing... I'll check and get back to it.

If all else fails, what about two packages: freetype2-bytecode freetype2-autohint where each provides=('freetype2')?
Comment by Aaron Griffin (phrakture) - Wednesday, 19 January 2005, 18:42 GMT
http://freetype.sourceforge.net/patents.html

There are also patent issues with the bytecode interpreter - technically, in order to use the version of freetype2 supplied by Arch, one must pay Apple...

here's some screenshots of the auto-hinter at work:
http://freetype.sourceforge.net/autohinting/hinter.html#screenshots

Judd, are your fonts anti-aliased in those screen-shots? That could be a issue as well.
Comment by Aaron Griffin (phrakture) - Friday, 21 January 2005, 17:06 GMT
Hmm, the auto-hinter default would work. I think. The bytecode interpreter was what was causing my issues, but I am not sure if the hangs will still happen due to some small chunk of code it hits before it realises that it should use the autohinter and not bytecode interp. Though JGC stated it had to do with bytecode caching...

Even scarier is the patent issue. I think there should be a disclaimer on install of this package and the mplayer codecs.
Comment by Eugenia Loli-Queru (Eugenia) - Thursday, 17 February 2005, 03:09 GMT
>XFFM with bytecode interpreter disabled (and AA/autohint enabled)

That's the best of the 3 attached screenshots. That's the default that it should be used by Arch. Both the other two first screenshots suck quality-wise.
Comment by Aaron Griffin (phrakture) - Thursday, 17 February 2005, 15:27 GMT
I agree, in addition, the Microsoft fonts do not work well with the bytecode interpreter - they are hinted, even on windows. (I think Apple does the same with their native fonts)
Comment by arjan timmerman (blaasvis) - Saturday, 05 March 2005, 17:41 GMT
with fontconfig-2.3.0-2 and the bytecode patch remove all bold fonts look horrible.....

i will add the screenshot.
Comment by James Rayner (iphitus) - Saturday, 05 March 2005, 23:48 GMT
I think the fonts look more awesome than ever with this. I use a lcd display.

maybe split and have two versions, one with, one without,

I havnt had any stability issues in any of my programs - I only use gtk2 programs though.
Comment by Aaron Griffin (phrakture) - Sunday, 06 March 2005, 02:41 GMT
arjan: I think those fonts look good.

james: I agree - two versions wuld be best
Comment by Simo Leone (neotuli) - Sunday, 06 March 2005, 17:33 GMT
My fonts get really ugly without bytecode, exactly as in the examples already shown.
Comment by Judd Vinet (judd) - Sunday, 06 March 2005, 20:11 GMT
It is very easy to enable/disable the bytecode interpreter.

Look at the first block in /etc/fonts/local.conf. It starts with a comment saying "Enable the auto-hinter". Comment that block out if you want to use the bytecode interpreter.

The jury's still out on what the Arch default should be, but two separate packages are not necessary. Stay tuned.
Comment by Simo Leone (neotuli) - Sunday, 03 April 2005, 02:38 GMT
[bug-squash day prelims]
I think this bug can close, since the new package clearly states how to turn it on and off, as well as which is the default, in its install-time echos.

Loading...