Core Text doesn't actually have a concept of DPI internally, as it
doesn't rasterize anything by itself, it just generates vector paths
that get passed along to Core Graphics.
In practice this means Core Text operates in the classical macOS
logical DPI of 72, with one typographic point corresponding to one
point in the Core Graphics coordinate system, which for a normal
bitmap context then corresponds to one pixel -- or two pixels for
a "retina" context with a 2x scale transform.
Scaling the font point sizes given to HarfBuzz to an assumed DPI
of 96 is problematic with this in mind, as fonts with optical
features such as 'trak' tables for tracking, or color glyphs,
will then base the metrics off of the wrong point size compared
to what the client asked for.
This in turn causes mismatches between the metrics of the shaped
text and the actual rasterization, which doesn't include the 72
to 96 DPI scaling.
If a 96 DPI is needed, such as on the Web, the scaling should be
done outside of HarfBuzz, allowing the client to keep the DPI of
the shaping in sync with the rasterization.
The recommended way to do that is by scaling the font point size,
not by applying a transform to the target Core Graphics context,
to let Core Text choose the right optical features of the target
point size, as described in WWDC 2015 session 804:
https://developer.apple.com/videos/play/wwdc2015/804/
It is not visually noticeable but apparently affected by kern format2 correct implementation.
I should've checked CoreText result which can't as CircleCI outage.
If we have duplicae font files in different directories, that would
break the oss-fuzz build currently. So, rename some to avoid
name class with text-rendering-test. Would be better to find
another solution.
The font supports the deprecated tag 'DHV ' instead of 'DIV '. dv is
mapped to 'DIV ' and 'DHV ', in that order. The test specifies
`--language=dv`, demonstrating that if a font does not support the first
OpenType tag mapped to a BCP 47 tag, it will fall back to the next tag.
Fixes https://github.com/harfbuzz/harfbuzz/issues/1019
New numbers:
BENGALI: 353725 out of 354188 tests passed. 463 failed (0.130722%)
DEVANAGARI: 707261 out of 707394 tests passed. 133 failed (0.0188014%)
GUJARATI: 366353 out of 366457 tests passed. 104 failed (0.0283799%)
GURMUKHI: 60729 out of 60747 tests passed. 18 failed (0.0296311%)
KANNADA: 951300 out of 951913 tests passed. 613 failed (0.0643966%)
MALAYALAM: 1048136 out of 1048334 tests passed. 198 failed (0.0188871%)
ORIYA: 42327 out of 42329 tests passed. 2 failed (0.00472489%)
SINHALA: 271596 out of 271847 tests passed. 251 failed (0.0923313%)
TAMIL: 1091754 out of 1091754 tests passed. 0 failed (0%)
TELUGU: 970555 out of 970573 tests passed. 18 failed (0.00185457%)
Devanagari regressed because Uniscribe doesn't enforce the full set.
Tests added with the *-vowel-letters.txt files in tree and Noto fonts.
Note that there's minor positioning differences, and ONE reordering
difference between what we get for these and what Uniscribe gets.
Probably same as what's described in commit message for
1a96cc825d
From the new code (first paragraph is from the OT Devanagari spec.):
/* o Reorder matras:
*
* If a pre-base matra character had been reordered before applying basic
* features, the glyph can be moved closer to the main consonant based on
* whether half-forms had been formed. Actual position for the matra is
* defined as “after last standalone halant glyph, after initial matra
* position and before the main consonant”. If ZWJ or ZWNJ follow this
* halant, position is moved after it.
*
* IMPLEMENTATION NOTES:
*
* It looks like the last sentence is wrong. Testing, with Windows 7 Uniscribe
* and Devanagari shows that the behavior is best described as:
*
* "If ZWJ follows this halant, matra is NOT repositioned after this halant.
* If ZWNJ follows this halant, position is moved after it."
*
* Test case, with Adobe Devanagari or Nirmala UI:
*
* U+091F,U+094D,U+200C,U+092F,U+093F
* (Matra moves to the middle, after ZWNJ.)
*
* U+091F,U+094D,U+200D,U+092F,U+093F
* (Matra does NOT move, stays to the left.)
Fixes https://github.com/harfbuzz/harfbuzz/issues/1070
Test case added with Adobe Devanagari.
David Corbett
Behdad Esfahbod
Ebrahim Byagowi
Tor Arne Vestbø
punchcutter
Khaled Hosny