more fixes

CONTRIBUTING.md

@@ -46,8 +46,8 @@ Rightmost vertical column (should be 20 px tall) contains the tags. Tags are def
       W  |= Width of the character
       W  |
       W -'
-      m --Is this character lowheight?
       K -,
+      K  |
       K  |= Tags used by the "Keming Machine"
       K -'
       Q ---Compiler Directive (see below)
@@ -77,29 +77,32 @@ Up&Down:
 
     <MSB,Red> SXXXXXXX SYYYYYYY 00000000 <LSB,Blue>
 
-Each X and Y numbers are Signed 8-Bit Integer.
+Each X and Y numbers are TWO'S COMPLEMENT Signed 8-Bit Integer.
 
 X-positive: nudges towards left  
-Y-positive: nudges towards up  
+Y-positive: nudges towards down  
 
 #### Diacritics Anchor Point Encoding
 
 4 Pixels are further divided as follows:
 
 | LSB |   | Red | Green | Blue |
-| ------------ | ------------ | ------------ | ------------ | ------------ |
+| ------------ | ------------ | ------------ | ----------- | ------------ |
 | Y | Anchor point Y for: | undefined | undefined | undefined |
 | X | Anchor point X for: | undefined | undefined | undefined |
-| Y | Anchor point Y for: | (unused) | (unused) | (unused) |
+| Y | Anchor point Y for: | Type-0 | Type-1 | Type-2 |
 | X | Anchor point X for: | Type-0 | Type-1 | Type-2 |
 | **MSB** |   |   |   |   |
 
-    <MSB,Red> 1Y1Y1Y1Y 1Y2Y2Y2Y 1Y3Y3Y3Y <LSB,Blue>
+    <MSB,Red> 1Y1Y1Y1Y 2Y2Y2Y2Y 3Y3Y3Y3Y <LSB,Blue>
-    <MSB,Red> 1X1X1X1X 1X2X2X2X 1X3X3X3X <LSB,Blue>
+    <MSB,Red> 1X1X1X1X 2X2X2X2X 3X3X3X3X <LSB,Blue>
 
 where Red is first, Green is second, Blue is the third diacritics.
-MSB for each word must be set so that the pixel would appear brighter on the image editor.
+
-(the font program will only read low 7 bits for each RGB channel)
+Each X and Y numbers are SIGN AND MAGNITUDE 8-Bit Integer.
+
+X-positive: nudges towards left  
+Y-positive: nudges towards down
 
 #### Diacritics Type Bit Encoding
 

OTFbuild/font_builder.py

@@ -370,10 +370,9 @@ def build_font(assets_dir, output_path, no_bitmap=False, no_features=False):
             x_offset = 0
         x_offset -= g.props.nudge_x * SCALE
 
-        # For STACK_DOWN marks (below-base diacritics), negative nudge_y
+        # For marks (write_on_top >= 0), positive nudge_y means shift UP
-        # means "shift content down to below baseline".  The sign convention
+        # in the bitmap engine (opposite to non-marks where positive = down).
-        # is opposite to non-marks where positive nudge_y means shift down.
+        if g.props.write_on_top >= 0:
-        if g.props.stack_where == SC.STACK_DOWN and g.props.write_on_top >= 0:
             y_offset = g.props.nudge_y * SCALE
         else:
             y_offset = -g.props.nudge_y * SCALE

OTFbuild/glyph_parser.py

@@ -38,6 +38,7 @@ class GlyphProps:
     has_kern_data: bool = False
     is_kern_y_type: bool = False
     kerning_mask: int = 255
+    dot_removal: Optional[int] = None  # codepoint to replace with when followed by a STACK_UP mark
     directive_opcode: int = 0
     directive_arg1: int = 0
     directive_arg2: int = 0
@@ -131,7 +132,8 @@ def parse_variable_sheet(image, sheet_index, cell_w, cell_h, cols, is_xy_swapped
 
         # Kerning data
         kerning_bit1 = _tagify(image.get_pixel(code_start_x, code_start_y + 6))
-        # kerning_bit2 and kerning_bit3 are reserved
+        kerning_bit2 = _tagify(image.get_pixel(code_start_x, code_start_y + 7))
+        dot_removal = None if kerning_bit2 == 0 else (kerning_bit2 >> 8)
         is_kern_y_type = (kerning_bit1 & 0x80000000) != 0
         kerning_mask = (kerning_bit1 >> 8) & 0xFFFFFF
         has_kern_data = (kerning_bit1 & 0xFF) != 0
@@ -188,7 +190,7 @@ def parse_variable_sheet(image, sheet_index, cell_w, cell_h, cols, is_xy_swapped
             align_where=align_where, write_on_top=write_on_top,
             stack_where=stack_where, ext_info=ext_info,
             has_kern_data=has_kern_data, is_kern_y_type=is_kern_y_type,
-            kerning_mask=kerning_mask,
+            kerning_mask=kerning_mask, dot_removal=dot_removal,
             directive_opcode=directive_opcode, directive_arg1=directive_arg1,
             directive_arg2=directive_arg2,
         )

OTFbuild/opentype_features.py

@@ -70,6 +70,11 @@ languagesystem sund dflt;
     if ccmp_code:
         parts.append(ccmp_code)
 
+    # ccmp: dot removal (e.g. i→ı, j→ȷ when followed by STACK_UP marks)
+    dot_removal_code = _generate_dot_removal(glyphs, has)
+    if dot_removal_code:
+        parts.append(dot_removal_code)
+
     # Hangul jamo GSUB assembly
     hangul_code = _generate_hangul_gsub(glyphs, has, jamo_data)
     if hangul_code:
@@ -162,6 +167,54 @@ def _generate_ccmp(replacewith_subs, has):
     return '\n'.join(lines)
 
 
+def _generate_dot_removal(glyphs, has):
+    """Generate ccmp contextual substitution for dot removal.
+
+    When a base glyph tagged with dot_removal (kerning bit 2, pixel Y+7) is
+    followed by a STACK_UP mark, substitute the base with its dotless form.
+    Matches the Kotlin engine's dotRemoval logic.
+    """
+    # Collect all STACK_UP marks
+    stack_up_marks = []
+    for cp, g in glyphs.items():
+        if g.props.write_on_top >= 0 and g.props.stack_where == SC.STACK_UP and has(cp):
+            stack_up_marks.append(cp)
+
+    if not stack_up_marks:
+        return ""
+
+    # Collect all base glyphs with dot_removal
+    dot_removal_subs = []
+    for cp, g in glyphs.items():
+        if g.props.dot_removal is not None and has(cp) and has(g.props.dot_removal):
+            dot_removal_subs.append((cp, g.props.dot_removal))
+
+    if not dot_removal_subs:
+        return ""
+
+    lines = []
+
+    # Define the STACK_UP marks class
+    mark_names = ' '.join(glyph_name(cp) for cp in sorted(stack_up_marks))
+    lines.append(f"@stackUpMarks = [{mark_names}];")
+    lines.append("")
+
+    # Single substitution lookup for the replacements
+    lines.append("lookup DotRemoval {")
+    for src_cp, dst_cp in sorted(dot_removal_subs):
+        lines.append(f"    sub {glyph_name(src_cp)} by {glyph_name(dst_cp)};")
+    lines.append("} DotRemoval;")
+    lines.append("")
+
+    # Contextual rules in ccmp
+    lines.append("feature ccmp {")
+    for src_cp, _ in sorted(dot_removal_subs):
+        lines.append(f"    sub {glyph_name(src_cp)}' lookup DotRemoval @stackUpMarks;")
+    lines.append("} ccmp;")
+
+    return '\n'.join(lines)
+
+
 def _generate_hangul_gsub(glyphs, has, jamo_data):
     """
     Generate Hangul jamo GSUB lookups for syllable assembly.
@@ -1750,7 +1803,7 @@ def _generate_mark(glyphs, has):
 
     mark_groups = {}  # (mark_type, align, is_dia, stack_cat) -> [(cp, g), ...]
     for cp, g in marks.items():
-        is_dia = (0x0300 <= cp <= 0x036F)
+        is_dia = True  # all marks (write_on_top >= 0) are diacritics; Kotlin applies lowheight shiftdown unconditionally
         sc = _stack_cat(g.props.stack_where)
         key = (g.props.write_on_top, g.props.align_where, is_dia, sc)
         mark_groups.setdefault(key, []).append((cp, g))
@@ -1846,7 +1899,9 @@ def _generate_mark(glyphs, has):
             # Lowheight adjustment for combining diacritical marks:
             # shift base anchor Y down so diacritics sit closer to
             # the shorter base glyph.
-            if is_dia and g.props.is_low_height:
+            # Only applies to 'up' marks (and overlay), not 'dn',
+            # matching Kotlin which only adjusts in STACK_UP/STACK_UP_N_DOWN.
+            if is_dia and g.props.is_low_height and scat != 'dn':
                 if mark_type == 2:  # overlay
                     ay -= SC.H_OVERLAY_LOWERCASE_SHIFTDOWN * SC.SCALE
                 else:  # above (type 0)
@@ -1878,12 +1933,13 @@ def _generate_mark(glyphs, has):
         lines.append(f"lookup {mkmk_name} {{")
 
         if scat == 'up':
-            m2y = SC.ASCENT + SC.H_DIACRITICS * SC.SCALE
+            m2y_base = SC.ASCENT + SC.H_DIACRITICS * SC.SCALE
         else:  # 'dn'
-            m2y = SC.ASCENT - SC.H_DIACRITICS * SC.SCALE
+            m2y_base = SC.ASCENT - SC.H_DIACRITICS * SC.SCALE
 
         for cp, g in mark_list:
             mx = mark_anchors.get(cp, 0)
+            m2y = m2y_base
             lines.append(
                 f"    pos mark {glyph_name(cp)}"
                 f" <anchor {mx} {m2y}> mark {class_name};"
@@ -1893,6 +1949,46 @@ def _generate_mark(glyphs, has):
         lines.append("")
         mkmk_lookup_names.append(mkmk_name)
 
+    # --- Nudge-Y gap correction for MarkToMark ---
+    # Without cascade, the gap between consecutive 'up' marks is
+    #   H_DIACRITICS + nudge_y_2 - nudge_y_1
+    # which is less than H_DIACRITICS when nudge_y_1 > nudge_y_2
+    # (e.g. Cyrillic uni2DED nudge=2 followed by uni0487 nudge=0).
+    # Add contextual positioning to compensate: shift mark2 up by
+    # (nudge_y_1 - nudge_y_2) * SCALE for each such pair.
+    # This keeps Thai correct (same nudge on both marks → no correction)
+    # while fixing Cyrillic (different nudge → correction applied).
+    nudge_groups = {}  # nudge_y -> [glyph_name, ...]
+    for (mark_type, align, is_dia, scat), mark_list in sorted(mark_groups.items()):
+        if scat != 'up':
+            continue
+        for cp, g in mark_list:
+            ny = g.props.nudge_y
+            nudge_groups.setdefault(ny, []).append(glyph_name(cp))
+
+    distinct_nudges = sorted(nudge_groups.keys())
+    correction_pairs = []
+    for n1 in distinct_nudges:
+        for n2 in distinct_nudges:
+            if n1 > n2:
+                correction_pairs.append((n1, n2, (n1 - n2) * SC.SCALE))
+
+    if correction_pairs:
+        for ny, glyphs in sorted(nudge_groups.items()):
+            lines.append(f"@up_nudge_{ny} = [{' '.join(sorted(glyphs))}];")
+        lines.append("")
+
+        mkmk_corr_name = "mkmk_nudge_correct"
+        lines.append(f"lookup {mkmk_corr_name} {{")
+        for n1, n2, val in correction_pairs:
+            lines.append(
+                f"    pos @up_nudge_{n1} <0 0 0 0>"
+                f" @up_nudge_{n2} <0 {val} 0 0>;"
+            )
+        lines.append(f"}} {mkmk_corr_name};")
+        lines.append("")
+        mkmk_lookup_names.append(mkmk_corr_name)
+
     # Register MarkToBase lookups under mark.
     # dev2 is excluded: HarfBuzz/DirectWrite use abvm for Devanagari marks.
     # deva is INCLUDED: CoreText's old-Indic shaper may need mark/mkmk

OTFbuild/sheet_config.py

@@ -72,6 +72,9 @@ SHEET_ALPHABETIC_PRESENTATION_FORMS = 38
 SHEET_HENTAIGANA_VARW = 39
 SHEET_CONTROL_PICTURES_VARW = 40
 SHEET_LEGACY_COMPUTING_VARW = 41
+SHEET_CYRILIC_EXTB_VARW = 42
+SHEET_CYRILIC_EXTA_VARW = 43
+SHEET_CYRILIC_EXTC_VARW = 44
 
 SHEET_UNKNOWN = 254
 
@@ -118,6 +121,9 @@ FILE_LIST = [
     "hentaigana_variable.tga",
     "control_pictures_variable.tga",
     "symbols_for_legacy_computing_variable.tga",
+    "cyrilic_extB_variable.tga",
+    "cyrilic_extA_variable.tga",
+    "cyrilic_extC_variable.tga",
 ]
 
 CODE_RANGE = [
@@ -151,7 +157,7 @@ CODE_RANGE = [
     list(range(0xA720, 0xA800)),                                                    # 27: Latin Ext D
     list(range(0x20A0, 0x20D0)),                                                    # 28: Currencies
     list(range(0xFFE00, 0xFFFA0)),                                                  # 29: Internal
-    list(range(0x2100, 0x2150)),                                                    # 30: Letterlike
+    list(range(0x2100, 0x2200)),                                                    # 30: Letterlike
     list(range(0x1F100, 0x1F200)),                                                  # 31: Enclosed Alphanum Supl
     list(range(0x0B80, 0x0C00)) + list(range(0xF00C0, 0xF0100)),                    # 32: Tamil
     list(range(0x980, 0xA00)),                                                      # 33: Bengali
@@ -163,6 +169,9 @@ CODE_RANGE = [
     list(range(0x1B000, 0x1B170)),                                                  # 39: Hentaigana
     list(range(0x2400, 0x2440)),                                                    # 40: Control Pictures
     list(range(0x1FB00, 0x1FC00)),                                                  # 41: Legacy Computing
+    list(range(0xA640, 0xA6A0)),                                                    # 42: Cyrillic Ext B
+    list(range(0x2DE0, 0x2E00)),                                                    # 43: Cyrillic Ext A
+    list(range(0x1C80, 0x1C8F)),                                                    # 43: Cyrillic Ext C
 ]
 
 CODE_RANGE_HANGUL_COMPAT = range(0x3130, 0x3190)
@@ -539,5 +548,8 @@ def index_y(sheet_index, c):
         SHEET_HENTAIGANA_VARW: lambda: (c - 0x1B000) // 16,
         SHEET_CONTROL_PICTURES_VARW: lambda: (c - 0x2400) // 16,
         SHEET_LEGACY_COMPUTING_VARW: lambda: (c - 0x1FB00) // 16,
+        SHEET_CYRILIC_EXTB_VARW: lambda: (c - 0xA640) // 16,
+        SHEET_CYRILIC_EXTA_VARW: lambda: (c - 0x2DE0) // 16,
+        SHEET_CYRILIC_EXTC_VARW: lambda: (c - 0x1C80) // 16,
         SHEET_HANGUL: lambda: 0,
     }.get(sheet_index, lambda: c // 16)()

demotext_unaligned.txt

@@ -25,6 +25,7 @@ How multilingual? Real multilingual!
 􏻬আমি কাঁচ খেতে পারি, তাতে আমার কোনো ক্ষতি হয় না। 􀀀
 􏻬󿿁Под южно дърво, цъфтящо в синьо, бягаше малко пухкаво зайче󿿀􀀀
 􏻬ᎠᏍᎦᏯᎡᎦᎢᎾᎨᎢᎣᏍᏓᎤᎩᏍᏗᎥᎴᏓᎯᎲᎢᏔᎵᏕᎦᏟᏗᏖᎸᎳᏗᏗᎧᎵᎢᏘᎴᎩ ᏙᏱᏗᏜᏫᏗᏣᏚᎦᏫᏛᏄᏓᎦᏝᏃᎠᎾᏗᎭᏞᎦᎯᎦᏘᏓᏠᎨᏏᏕᏡᎬᏢᏓᏥᏩᏝᎡᎢᎪᎢ ᎠᎦᏂᏗᎮᎢᎫᎩᎬᏩᎴᎢᎠᏆᏅᏛᎫᏊᎾᎥᎠᏁᏙᎲᏐᏈᎵᎤᎩᎸᏓᏭᎷᏤᎢᏏᏉᏯᏌᏊ ᎤᏂᏋᎢᏡᎬᎢᎰᏩᎬᏤᎵᏍᏗᏱᎩᎱᎱᎤᎩᎴᎢᏦᎢᎠᏂᏧᏣᏨᎦᏥᎪᎥᏌᏊᎤᎶᏒᎢᎢᏡᎬᎢ ᎹᎦᎺᎵᏥᎻᎼᏏᎽᏗᏩᏂᎦᏘᎾᎿᎠᏁᎬᎢᏅᎩᎾᏂᎡᎢᏌᎶᎵᏎᎷᎠᏑᏍᏗᏪᎩ ᎠᎴ ᏬᏗᏲᏭᎾᏓᏍᏓᏴᏁᎢᎤᎦᏅᏮᏰᎵᏳᏂᎨᎢ􀀀
+􏻬Ѳеѡфа́нъ и҆ Алеѯі́й, ѕѣлѡ̀ возлюби́вше ѱалти́рь, воспѣ́ша при свѣ́тѣ ѕвѣ́здъ, помазꙋ́юще сщ҃е́нное мѵ́ро; серафими мн̑оꙮ҆читїи̑, ꙗ҆́кѡ ѻ҆́гнь, ѡ҆крꙋжа́хꙋ прⷭ҇то́лъ Бж҃їй, и҆ всѧ̀ землѧ̀ и҆спо́лнисѧ свѣ́та, ꙗ҆́кѡ ѕмі́й попра́нъ є҆́сть􀀀
 􏻬Příliš žluťoučký kůň úpěl ďábelské ódy􀀀
 􏻬Quizdeltagerne spiste jordbær med fløde, mens cirkusklovnen Walther spillede på xylofon􀀀
 􏻬PACK MY BOX WITH FIVE DOZEN LIQUOR JUGS􀀀
@@ -104,6 +105,10 @@ How multilingual? Real multilingual!
 
 􎳌‣ Full support for Archaic Kana/Hentaigana􀀀
 
+􏻬серафими мн̑оꙮ҆читїи̑, ꙗ҆́кѡ ѻ҆́гнь, ѡ҆крꙋжа́хꙋ прⷭ҇то́лъ Бж҃їй, и҆ всѧ̀ землѧ̀ и҆спо́лнисѧ свѣ́та􀀀
+
+􎳌‣ Fan of Church Slavonic? We’ve got you!􀀀
+
 􏃯Supported Unicode Blocks:􀀀
 
 ⁃ Basic Latin
@@ -111,6 +116,7 @@ How multilingual? Real multilingual!
 ⁃ Latin Extended Additional
 ⁃ Latin Extended-A/B/C/D
 ⁃ Armenian
+⁃ Arrows
 ⁃ Bengali􏿆ᶠⁱ􀀀
 ⁃ Braille Patterns
 ⁃ Cherokee􏿆⁷􀀀
@@ -120,8 +126,9 @@ How multilingual? Real multilingual!
 ⁃ Combining Diacritical Marks
 ⁃ Control Pictures
 ⁃ Currency Symbols
-⁃ Cyrillic􏿆ᴭ􀀀
+⁃ Cyrillic
-⁃ Cyrillic Supplement􏿆ᴭ􀀀
+⁃ Cyrillic Supplement
+⁃ Cyrillic Extended-A/B/C
 ⁃ Devanagari
 ⁃ Enclosed Alphanumeric Supplement
 ⁃ General Punctuations
@@ -140,6 +147,7 @@ How multilingual? Real multilingual!
 ⁃ Katakana Phonetic Extensions
 ⁃ Kana Supplement
 ⁃ Kana Extended-A
+⁃ Number Forms
 ⁃ Small Kana Extension
 ⁃ Letterlike Symbols
 ⁃ Phonetic Extensions
@@ -153,7 +161,7 @@ How multilingual? Real multilingual!
 ⁃ Tamil
 ⁃ Thai
 
-  􏿆ᴭ􀀀  No support for archæic letters　　􏿆ᴱ􀀀  No support for Coptic
+  􏿆ᴱ􀀀  No support for Coptic
   􏿆ᶠⁱ􀀀  No support for ligatures　　􏿆ჼ􀀀  Mkhedruli only
   􏿆⁶􀀀  􏿆⁷􀀀  􏿆⁹􀀀  􏿆¹²·¹􀀀  Up to the specified Unicode version
 

keming_machine.txt

@@ -1,7 +1,7 @@
 --- Pixel 0
 - Lowheight bit
 - encoding: has pixel - it's low height
-- used by the diacritics system to quickly look up if the character is low height without parsing the Pixel 1
+- bit must be set if above-diacritics should be lowered (e.g. lowercase b, which has 'A' shape bit but considered lowheight) 
 
 ### Legends
 #
@@ -106,3 +106,5 @@ dot removal for diacritics:
 - encoding:
     - <MSB> RRRRRRRR GGGGGGGG BBBBBBBB <LSB>
 
+--- Pixel 3
+Unused for now.

src/net/torvald/terrarumsansbitmap/GlyphProps.kt

@@ -1,9 +1,13 @@
 package net.torvald.terrarumsansbitmap
 
+import net.torvald.terrarumsansbitmap.gdx.CodePoint
+
 /**
  * Created by minjaesong on 2021-11-25.
  */
-data class DiacriticsAnchor(val type: Int, val x: Int, val y: Int, val xUsed: Boolean, val yUsed: Boolean)
+data class DiacriticsAnchor(val type: Int, val x: Int, val y: Int) {
+    val isZero = (x == 0 && y == 0)
+}
 /**
  * Created by minjaesong on 2018-08-07.
  */
@@ -15,7 +19,7 @@ data class GlyphProps(
         val nudgeX: Int = 0,
         val nudgeY: Int = 0,
 
-        val diacriticsAnchors: Array<DiacriticsAnchor> = Array(6) { DiacriticsAnchor(it, 0, 0, false, false) },
+        val diacriticsAnchors: Array<DiacriticsAnchor> = Array(6) { DiacriticsAnchor(it, 0, 0) },
 
         val alignWhere: Int = 0, // ALIGN_LEFT..ALIGN_BEFORE
 
@@ -29,6 +33,8 @@ data class GlyphProps(
         val isKernYtype: Boolean = false,
         val kerningMask: Int = 255,
 
+        val dotRemoval: CodePoint? = null,
+
         val directiveOpcode: Int = 0, // 8-bits wide
         val directiveArg1: Int = 0, // 8-bits wide
         val directiveArg2: Int = 0, // 8-bits wide
@@ -95,10 +101,6 @@ data class GlyphProps(
         diacriticsAnchors.forEach {
             hash = hash xor it.type
             hash = hash * 16777619
-            hash = hash xor (it.x or (if (it.xUsed) 128 else 0))
-            hash = hash * 16777619
-            hash = hash xor (it.y or (if (it.yUsed) 128 else 0))
-            hash = hash * 16777619
         }
 
         hash = hash xor tags

src/net/torvald/terrarumsansbitmap/gdx/TerrarumSansBitmap.kt

@@ -888,6 +888,9 @@ class TerrarumSansBitmap(
             SHEET_HENTAIGANA_VARW -> hentaiganaIndexY(ch)
             SHEET_CONTROL_PICTURES_VARW -> controlPicturesIndexY(ch)
             SHEET_LEGACY_COMPUTING_VARW -> legacyComputingIndexY(ch)
+            SHEET_CYRILIC_EXTB_VARW -> cyrilicExtBIndexY(ch)
+            SHEET_CYRILIC_EXTA_VARW -> cyrilicExtAIndexY(ch)
+            SHEET_CYRILIC_EXTC_VARW -> cyrilicExtCIndexY(ch)
             else -> ch / 16
         }
 
@@ -918,9 +921,9 @@ class TerrarumSansBitmap(
             val isLowHeight = (pixmap.getPixel(codeStartX, codeStartY + 5).and(255) != 0)
 
             // Keming machine parameters
-            val kerningBit1 = pixmap.getPixel(codeStartX, codeStartY + 6).tagify()
+            val kerningBit1 = pixmap.getPixel(codeStartX, codeStartY + 6).tagify() // glyph shape
-            val kerningBit2 = pixmap.getPixel(codeStartX, codeStartY + 7).tagify()
+            val kerningBit2 = pixmap.getPixel(codeStartX, codeStartY + 7).tagify() // dot removal
-            val kerningBit3 = pixmap.getPixel(codeStartX, codeStartY + 8).tagify()
+            val kerningBit3 = pixmap.getPixel(codeStartX, codeStartY + 8).tagify() // unused
             var isKernYtype = ((kerningBit1 and 0x80000000.toInt()) != 0)
             var kerningMask = kerningBit1.ushr(8).and(0xFFFFFF)
             val hasKernData = kerningBit1 and 255 != 0//(kerningBit1 and 255 != 0 && kerningMask != 0xFFFF)
@@ -944,12 +947,12 @@ class TerrarumSansBitmap(
                 val shift = (3 - (it % 3)) * 8
                 val yPixel = pixmap.getPixel(codeStartX, codeStartY + yPos).tagify()
                 val xPixel = pixmap.getPixel(codeStartX, codeStartY + yPos + 1).tagify()
-                val yUsed = (yPixel ushr shift) and 128 != 0
+                val ySgn = ((yPixel ushr shift) and 128).let { if (it == 0) -1 else 1 }
-                val xUsed = (xPixel ushr shift) and 128 != 0
+                val xSgn = ((xPixel ushr shift) and 128).let { if (it == 0) -1 else 1 }
-                val y = if (yUsed) (yPixel ushr shift) and 127 else 0
+                val y = ((yPixel ushr shift) and 127) * ySgn
-                val x = if (xUsed) (xPixel ushr shift) and 127 else 0
+                val x = ((xPixel ushr shift) and 127) * xSgn
 
-                DiacriticsAnchor(it, x, y, xUsed, yUsed)
+                DiacriticsAnchor(it, x, y)
             }.toTypedArray()
 
             val alignWhere = (0..1).fold(0) { acc, y -> acc or ((pixmap.getPixel(codeStartX, codeStartY + y + 15).and(255) != 0).toInt() shl y) }
@@ -968,7 +971,9 @@ class TerrarumSansBitmap(
                 GlyphProps.STACK_DONT
             else (0..1).fold(0) { acc, y -> acc or ((pixmap.getPixel(codeStartX, codeStartY + y + 18).and(255) != 0).toInt() shl y) }
 
-            glyphProps[code] = GlyphProps(width, isLowHeight, nudgeX, nudgeY, diacriticsAnchors, alignWhere, writeOnTop, stackWhere, IntArray(15), hasKernData, isKernYtype, kerningMask, directiveOpcode, directiveArg1, directiveArg2)
+            val dotRemoval = if (kerningBit2 == 0) null else kerningBit2.ushr(8)
+
+            glyphProps[code] = GlyphProps(width, isLowHeight, nudgeX, nudgeY, diacriticsAnchors, alignWhere, writeOnTop, stackWhere, IntArray(15), hasKernData, isKernYtype, kerningMask, dotRemoval, directiveOpcode, directiveArg1, directiveArg2)
 
             // extra info
             val extCount = glyphProps[code]?.requiredExtInfoCount() ?: 0
@@ -1015,17 +1020,6 @@ class TerrarumSansBitmap(
         (0xD800..0xDFFF).forEach { glyphProps[it] = GlyphProps(0) }
         (0x100000..0x10FFFF).forEach { glyphProps[it] = GlyphProps(0) }
         (0xFFFA0..0xFFFFF).forEach { glyphProps[it] = GlyphProps(0) }
-
-
-        // manually add width of one orphan insular letter
-        // WARNING: glyphs in 0xA770..0xA778 has invalid data, further care is required
-        glyphProps[0x1D79] = GlyphProps(9)
-
-
-        // U+007F is DEL originally, but this font stores bitmap of Replacement Character (U+FFFD)
-        // to this position. String replacer will replace U+FFFD into U+007F.
-        glyphProps[0x7F] = GlyphProps(15)
-
     }
 
     private fun Int.halveWidth() = this / 2 + 1
@@ -1120,6 +1114,7 @@ class TerrarumSansBitmap(
         var nonDiacriticCounter = 0 // index of last instance of non-diacritic char
         var stackUpwardCounter = 0 // TODO separate stack counter for centre- and right aligned
         var stackDownwardCounter = 0
+        var nudgeUpHighWater = 0 // tracks max nudgeY seen in current stack, so subsequent marks with lower nudge still clear the previous one
 
         val HALF_VAR_INIT = W_VAR_INIT.minus(1).div(2)
 
@@ -1198,6 +1193,7 @@ class TerrarumSansBitmap(
 
                         stackUpwardCounter = 0
                         stackDownwardCounter = 0
+                        nudgeUpHighWater = 0
                     }
                     // FIXME HACK: using 0th diacritics' X-anchor pos as a type selector
                     /*else if (thisProp.writeOnTop && thisProp.diacriticsAnchors[0].x == GlyphProps.DIA_JOINER) {
@@ -1217,30 +1213,32 @@ class TerrarumSansBitmap(
                         // set X pos according to alignment information
                         posXbuffer[charIndex] = -thisProp.nudgeX +
                                 when (thisProp.alignWhere) {
-                                    GlyphProps.ALIGN_LEFT, GlyphProps.ALIGN_BEFORE -> posXbuffer[nonDiacriticCounter]
+                                    GlyphProps.ALIGN_LEFT, GlyphProps.ALIGN_BEFORE -> {
+                                        val anchorPointX = if (itsProp.diacriticsAnchors[diacriticsType].isZero) itsProp.width else itsProp.diacriticsAnchors[diacriticsType].x
+
+                                        posXbuffer[nonDiacriticCounter] + anchorPointX
+                                    }
                                     GlyphProps.ALIGN_RIGHT -> {
 //                                        println("thisprop alignright $kerning, $extraWidth")
 
-                                        val anchorPoint =
+                                        val anchorPointX = if (itsProp.diacriticsAnchors[diacriticsType].isZero) itsProp.width else itsProp.diacriticsAnchors[diacriticsType].x
-                                            if (!itsProp.diacriticsAnchors[diacriticsType].xUsed) itsProp.width else itsProp.diacriticsAnchors[diacriticsType].x
 
                                         extraWidth += thisProp.width
-                                        posXbuffer[nonDiacriticCounter] + anchorPoint - W_VAR_INIT + kerning + extraWidth
+                                        posXbuffer[nonDiacriticCounter] + anchorPointX - W_VAR_INIT + kerning + extraWidth
                                     }
                                     GlyphProps.ALIGN_CENTRE -> {
-                                        val anchorPoint =
+                                        val anchorPointX = if (itsProp.diacriticsAnchors[diacriticsType].isZero) itsProp.width.div(2) else itsProp.diacriticsAnchors[diacriticsType].x
-                                            if (!itsProp.diacriticsAnchors[diacriticsType].xUsed) itsProp.width.div(2) else itsProp.diacriticsAnchors[diacriticsType].x
 
                                         if (itsProp.alignWhere == GlyphProps.ALIGN_RIGHT) {
                                             if (thisChar in 0x900..0x902)
-                                                posXbuffer[nonDiacriticCounter] + anchorPoint + (itsProp.width - 1).div(2)
+                                                posXbuffer[nonDiacriticCounter] + anchorPointX + (itsProp.width - 1).div(2)
                                             else
-                                                posXbuffer[nonDiacriticCounter] + anchorPoint + (itsProp.width + 1).div(2)
+                                                posXbuffer[nonDiacriticCounter] + anchorPointX + (itsProp.width + 1).div(2)
                                         } else {
                                             if (thisChar in 0x900..0x902)
-                                                posXbuffer[nonDiacriticCounter] + anchorPoint - (W_VAR_INIT + 1) / 2
+                                                posXbuffer[nonDiacriticCounter] + anchorPointX - (W_VAR_INIT + 1) / 2
                                             else
-                                                posXbuffer[nonDiacriticCounter] + anchorPoint - HALF_VAR_INIT
+                                                posXbuffer[nonDiacriticCounter] + anchorPointX - HALF_VAR_INIT
                                         }
                                     }
                                     else -> throw InternalError("Unsupported alignment: ${thisProp.alignWhere}")
@@ -1277,14 +1275,14 @@ class TerrarumSansBitmap(
                         // set Y pos according to diacritics position
                         when (thisProp.stackWhere) {
                             GlyphProps.STACK_DOWN -> {
-                                posYbuffer[charIndex] = (-thisProp.nudgeY + H_DIACRITICS * stackDownwardCounter) * flipY.toSign()
+                                posYbuffer[charIndex] = (-thisProp.nudgeY + H_DIACRITICS * stackDownwardCounter) * flipY.toSign() - thisProp.nudgeY
                                 stackDownwardCounter++
                             }
                             GlyphProps.STACK_UP -> {
-                                posYbuffer[charIndex] = -thisProp.nudgeY + (-H_DIACRITICS * stackUpwardCounter + -thisProp.nudgeY) * flipY.toSign()
+                                val effectiveNudge = maxOf(thisProp.nudgeY, nudgeUpHighWater)
+                                posYbuffer[charIndex] = -effectiveNudge + (-H_DIACRITICS * stackUpwardCounter + -effectiveNudge) * flipY.toSign() + effectiveNudge
                                 // shift down on lowercase if applicable
-                                if (getSheetType(thisChar) in autoShiftDownOnLowercase &&
+                                if (lastNonDiacriticChar.isLowHeight()) {
-                                    lastNonDiacriticChar.isLowHeight()) {
                                     //dbgprn("AAARRRRHHHH for character ${thisChar.toHex()}")
                                     //dbgprn("lastNonDiacriticChar: ${lastNonDiacriticChar.toHex()}")
                                     //dbgprn("cond: ${thisProp.alignXPos == GlyphProps.DIA_OVERLAY}, charIndex: $charIndex")
@@ -1294,6 +1292,7 @@ class TerrarumSansBitmap(
                                         posYbuffer[charIndex] += H_STACKUP_LOWERCASE_SHIFTDOWN * flipY.toSign() // if minus-assign doesn't work, try plus-assign
                                 }
 
+                                nudgeUpHighWater = effectiveNudge
                                 stackUpwardCounter++
 
 //                                    dbgprn("lastNonDiacriticChar: ${lastNonDiacriticChar.charInfo()}; stack counter: $stackUpwardCounter")
@@ -1302,22 +1301,25 @@ class TerrarumSansBitmap(
                                 posYbuffer[charIndex] = (-thisProp.nudgeY + H_DIACRITICS * stackDownwardCounter) * flipY.toSign()
                                 stackDownwardCounter++
 
-
+                                val effectiveNudge = maxOf(thisProp.nudgeY, nudgeUpHighWater)
-                                posYbuffer[charIndex] = (-thisProp.nudgeY + -H_DIACRITICS * stackUpwardCounter) * flipY.toSign()
+                                posYbuffer[charIndex] = (-effectiveNudge + -H_DIACRITICS * stackUpwardCounter) * flipY.toSign()
                                 // shift down on lowercase if applicable
-                                if (getSheetType(thisChar) in autoShiftDownOnLowercase &&
+                                if (lastNonDiacriticChar.isLowHeight()) {
-                                    lastNonDiacriticChar.isLowHeight()) {
                                     if (diacriticsType == GlyphProps.DIA_OVERLAY)
                                         posYbuffer[charIndex] += H_OVERLAY_LOWERCASE_SHIFTDOWN * flipY.toSign() // if minus-assign doesn't work, try plus-assign
                                     else
                                         posYbuffer[charIndex] += H_STACKUP_LOWERCASE_SHIFTDOWN * flipY.toSign() // if minus-assign doesn't work, try plus-assign
                                 }
 
+                                nudgeUpHighWater = effectiveNudge
                                 stackUpwardCounter++
                             }
                             // for BEFORE_N_AFTER, do nothing in here
                         }
 
+                        // nudge Y pos according to anchor position
+                        posYbuffer[charIndex] -= itsProp.diacriticsAnchors[diacriticsType].y
+
                         // Don't reset extraWidth here!
                     }
                 }
@@ -1327,7 +1329,7 @@ class TerrarumSansBitmap(
             if (str.isNotEmpty()) {
                 val lastCharProp = glyphProps[str.last()]
                 val penultCharProp = glyphProps[str[nonDiacriticCounter]] ?:
-                (if (errorOnUnknownChar) throw throw InternalError("No GlyphProps for char '${str[nonDiacriticCounter]}' " +
+                (if (errorOnUnknownChar) throw InternalError("No GlyphProps for char '${str[nonDiacriticCounter]}' " +
                         "(${str[nonDiacriticCounter].charInfo()})") else nullProp)
                 posXbuffer[posXbuffer.lastIndex] = posXbuffer[posXbuffer.lastIndex - 1] + // DON'T add 1 to house the shadow, it totally breaks stuffs
                         if (lastCharProp != null && lastCharProp.writeOnTop >= 0) {
@@ -1523,8 +1525,8 @@ class TerrarumSansBitmap(
 
             }
             // for lowercase i and j, if cNext is a diacritic that goes on top, remove the dots
-            else if (diacriticDotRemoval.containsKey(c) && (glyphProps[cNext]?.writeOnTop ?: -1) >= 0 && glyphProps[cNext]?.stackWhere == GlyphProps.STACK_UP) {
+            else if (glyphProps[c]!!.dotRemoval != null && (glyphProps[cNext]?.writeOnTop ?: -1) >= 0 && glyphProps[cNext]?.stackWhere == GlyphProps.STACK_UP) {
-                seq.add(diacriticDotRemoval[c]!!)
+                seq.add(glyphProps[c]!!.dotRemoval!!)
             }
 
             // BEGIN of tamil subsystem implementation
@@ -2604,6 +2606,9 @@ class TerrarumSansBitmap(
         internal const val SHEET_HENTAIGANA_VARW = 39
         internal const val SHEET_CONTROL_PICTURES_VARW = 40
         internal const val SHEET_LEGACY_COMPUTING_VARW = 41
+        internal const val SHEET_CYRILIC_EXTB_VARW = 42
+        internal const val SHEET_CYRILIC_EXTA_VARW = 43
+        internal const val SHEET_CYRILIC_EXTC_VARW = 44
 
         internal const val SHEET_UNKNOWN = 254
 
@@ -2625,10 +2630,6 @@ class TerrarumSansBitmap(
         const val MOVABLE_BLOCK_1 = 0xFFFF0
 
 
-        private val autoShiftDownOnLowercase = arrayOf(
-            SHEET_DIACRITICAL_MARKS_VARW
-        )
-
         private val fileList = arrayOf( // MUST BE MATCHING WITH SHEET INDICES!!
             "ascii_variable.tga",
             "hangul_johab.tga",
@@ -2672,6 +2673,9 @@ class TerrarumSansBitmap(
             "hentaigana_variable.tga",
             "control_pictures_variable.tga",
             "symbols_for_legacy_computing_variable.tga",
+            "cyrilic_extB_variable.tga",
+            "cyrilic_extA_variable.tga",
+            "cyrilic_extC_variable.tga",
         )
         internal val codeRange = arrayOf( // MUST BE MATCHING WITH SHEET INDICES!!
             0..0xFF, // SHEET_ASCII_VARW
@@ -2704,7 +2708,7 @@ class TerrarumSansBitmap(
             0xA720..0xA7FF, // SHEET_EXTD_VARW
             0x20A0..0x20CF, // SHEET_CURRENCIES_VARW
             0xFFE00..0xFFF9F, // SHEET_INTERNAL_VARW
-            0x2100..0x214F, // SHEET_LETTERLIKE_MATHS_VARW
+            0x2100..0x21FF, // SHEET_LETTERLIKE_MATHS_VARW
             0x1F100..0x1F1FF, // SHEET_ENCLOSED_ALPHNUM_SUPL_VARW
             (0x0B80..0x0BFF) + (0xF00C0..0xF00FF), // SHEET_TAMIL_VARW
             0x980..0x9FF, // SHEET_BENGALI_VARW
@@ -2716,6 +2720,9 @@ class TerrarumSansBitmap(
             0x1B000..0x1B16F, // SHEET_HENTAIGANA_VARW
             0x2400..0x243F, // SHEET_CONTROL_PICTURES_VARW
             0x1FB00..0x1FBFF, // SHEET_LEGACY_COMPUTING_VARW
+            0xA640..0xA69F, // SHEET_CYRILIC_EXTB_VARW
+            0x2DE0..0x2DFF, // SHEET_CYRILIC_EXTA_VARW
+            0x1C80..0x1C8F, // SHEET_CYRILIC_EXTC_VARW
         )
         private val codeRangeHangulCompat = 0x3130..0x318F
 
@@ -2733,11 +2740,6 @@ class TerrarumSansBitmap(
                 0x20..0x7F,
         )
 
-        private val diacriticDotRemoval = hashMapOf(
-            'i'.toInt() to 0x131,
-            'j'.toInt() to 0x237
-        )
-
         internal fun Int.charInfo() = "U+${this.toString(16).padStart(4, '0').toUpperCase()}: ${Character.getName(this)}"
 
         const val NQSP = 0x2000
@@ -3066,6 +3068,9 @@ class TerrarumSansBitmap(
         private fun hentaiganaIndexY(c: CodePoint) = (c - 0x1B000) / 16
         private fun controlPicturesIndexY(c: CodePoint) = (c - 0x2400) / 16
         private fun legacyComputingIndexY(c: CodePoint) = (c - 0x1FB00) / 16
+        private fun cyrilicExtBIndexY(c: CodePoint) = (c - 0xA640) / 16
+        private fun cyrilicExtAIndexY(c: CodePoint) = (c - 0x2DE0) / 16
+        private fun cyrilicExtCIndexY(c: CodePoint) = (c - 0x1C80) / 16
 
         val charsetOverrideDefault = Character.toChars(CHARSET_OVERRIDE_DEFAULT).toSurrogatedString()
         val charsetOverrideBulgarian = Character.toChars(CHARSET_OVERRIDE_BG_BG).toSurrogatedString()

src/net/torvald/terrarumtypewriterbitmap/gdx/TerrarumTypewriterBitmap.kt

@@ -229,16 +229,16 @@ class TerrarumTypewriterBitmap(
             val nudgeY = nudgingBits.ushr(16).toByte().toInt() // signed 8-bit int
 
             val diacriticsAnchors = (0..5).map {
-                val yPos = 11 + (it / 3) * 2
+                val yPos = 13 - (it / 3) * 2
                 val shift = (3 - (it % 3)) * 8
                 val yPixel = pixmap.getPixel(codeStartX, codeStartY + yPos).tagify()
                 val xPixel = pixmap.getPixel(codeStartX, codeStartY + yPos + 1).tagify()
-                val y = (yPixel ushr shift) and 127
+                val ySgn = ((yPixel ushr shift) and 128).let { if (it == 0) -1 else 1 }
-                val x = (xPixel ushr shift) and 127
+                val xSgn = ((xPixel ushr shift) and 128).let { if (it == 0) -1 else 1 }
-                val yUsed = (yPixel ushr shift) >= 128
+                val y = ((yPixel ushr shift) and 127) * ySgn
-                val xUsed = (yPixel ushr shift) >= 128
+                val x = ((xPixel ushr shift) and 127) * xSgn
 
-                DiacriticsAnchor(it, x, y, xUsed, yUsed)
+                DiacriticsAnchor(it, x, y)
             }.toTypedArray()
 
             val alignWhere = (0..1).fold(0) { acc, y -> acc or ((pixmap.getPixel(codeStartX, codeStartY + y + 15).and(255) != 0).toInt() shl y) }