Thu 17 Nov 2016 07:11:17 AM UTC, comment #8:
If you are going to use strong hinting, you must apply side bearing correction (i.e., the fields `lsb_delta' and `rsb_delta' in `FT_GlyphSlot') to compensate the increased side bearings that are intentionally added by the auto-hinter.
The attached ftstring images `without_delta.png' and `with_delta.png' showing `arialbd.ttf' at 10pt@96dpi neatly demonstrate this.
Note that it doesn't harm if the necessary code (as explained in the `FT_GlyphSlot' documentation) gets applied to any other hinting mode; the delta correction values are simply zero then.
This is a documentation bug. I've added a comment to the source code to explain why `pp1.x' is not needed. I've also improved the documentation of the delta values.
(file #38997, file #38998)
|
Wed 02 Nov 2016 02:29:52 PM UTC, comment #7:
This code there is really suspect:
pp1x_uh does not even depend on loader->pp1.x,
it is simply the movement of the left edge
whereas
pp2x_h is loader->pp1.x plus the movement of the right edge.
This is confusing.
|
Tue 12 Jun 2007 09:14:55 AM UTC, comment #6:
Could you please re-open this bug report?
It's still not fixed as of CVS 2007-06-12.
The spacing is really ugly on smaller font sizes, especially with bold fonts, so I have to patch freetype every time I download it. Having used my patched variant for several months now, I haven't noticed any problems yet (96dpi, full autohinting).
Thanks.
|
Thu 12 Apr 2007 03:13:43 PM UTC, comment #5:
>>Do you still think that this is necessary with the
>>current CVS version?
Yes, absolutely. Just tested with CVS from 2007-04-12, and I get exactly the same results as I described earlier. Patching it fixes the problem.
|
Mon 09 Apr 2007 08:18:50 PM UTC, comment #4:
Do you still think that this is necessary with the
current CVS version?
|
Wed 21 Feb 2007 07:22:29 PM UTC, comment #3:
After about half an hour of random torturing of the freetype source code, I finally found what was causing the distance to be so large. While I have absolutely no idea why it worked or how I found it, I'm pretty sure the patch fixes the "distance too large" problem for Arial 8pt bold @ 96dpi.
If anyone would enlighten me how this code works, I would be grateful. Thanks!
All this patch does is that it comments out 4 lines in afloader.c. Commenting only the first two or the second two lines didn't lead to the same good result.
Here's the patch:
diff -urNa freetype-2007-02-21/src/autofit/afloader.c freetype-2007-02-21.patched/src/autofit/afloader.c
--- freetype-2007-02-21/src/autofit/afloader.c 2007-02-13 01:28:21.000000000 +0400
+++ freetype-2007-02-21.patched/src/autofit/afloader.c 2007-02-21 23:02:33.000000000 +0400
@@ -206,12 +206,13 @@
loader->pp1.x = FT_PIX_ROUND( pp1x_uh );
loader->pp2.x = FT_PIX_ROUND( pp2x_uh );
+/*
if ( loader->pp1.x >= new_lsb && old_lsb > 0 )
loader->pp1.x -= 64;
if ( loader->pp2.x <= edge2->pos && old_rsb > 0 )
loader->pp2.x += 64;
-
+*/
slot->lsb_delta = loader->pp1.x - pp1x_uh;
slot->rsb_delta = loader->pp2.x - pp2x_uh;
}
|
Sat 17 Feb 2007 01:13:59 PM UTC, comment #2:
Thanks, the inconsistency seems to be fixed (as of cvs 2007-02-16) - the font looks much better now.
However, I think the spacing is still too wide, especially apparent with Arial 8pt bold font.
I'm attaching the composite screenshot of 4 modes - Arial 8pt (normal and bold) @ 96dpi, both freetype 2.1.10 and cvs 2007-02-14. (the screenshot says it's 2007-02-14, but it's the same with 2007-02-16).
I think this can also be reproduced with
ftview -r 96 8 arial.ttf
pressing "f" clearly shows that autohinted font is much wider (that is, the space between the characters is too wide).
Thank you for fixing the previous regressions.
(file #11994)
|
Wed 14 Feb 2007 11:41:25 PM UTC, comment #1:
I believe that this bug is now fixed in the CVS. Please test and I'll close it
|
Mon 12 Feb 2007 06:37:17 PM UTC, original submission:
Arial rendering has suffered a regression from 2.1.10 to 2.3.1 with autohinting on.
The spacing between the glyphs is really inconsistent in 2.3.1.
This issue has been reported as a part of a bug (overall Arial font rendering regression) at Novell bugzilla and has been confirmed.
Please, see the original bugreport at http://bugzilla.novell.com/show_bug.cgi?id=243558 (which also deals with a rendering glitch of glyph '9').
The screenshot that shows it best is available at http://bugzilla.novell.com/attachment.cgi?id=118081&action=view (freetype 2.1.10, 2.2.1 and 2.3.1 compared, Arial @ 10pt, autohint on, AA on, 96 dpi). Note the uneven spacing in 2.3.1.
The screenshot was taken using kwrite, but can also be replicated using ftview. However, ftview from 2.1.10 doesn't seem to accept a custom text string to display, so it's difficult to compare the two ftview outputs (the problem is not that apparent on a character table that ftview displays by default).
I can upload some ftview screenshots too if needed.
|