1gxvalid: TrueType GX validator
2==============================
3
4
51. What is this
6---------------
7
8 `gxvalid' is a module to validate TrueType GX tables: a collection of
9 additional tables in TrueType font which are used by `QuickDraw GX
10 Text', Apple Advanced Typography (AAT). In addition, gxvalid can
11 validates `kern' tables which have been extended for AAT. Like the
12 otvalid module, gxvalid uses FreeType 2's validator framework
13 (ftvalid).
14
15 You can link gxvalid with your program; before running your own layout
16 engine, gxvalid validates a font file. As the result, you can remove
17 error-checking code from the layout engine. It is also possible to
18 use gxvalid as a stand-alone font validator; the `ftvalid' test
19 program included in the ft2demo bundle calls gxvalid internally.
20 A stand-alone font validator may be useful for font developers.
21
22 This documents documents the following issues.
23
24 - supported TrueType GX tables
25 - fundamental validation limitations
26 - permissive error handling of broken GX tables
27 - `kern' table issue.
28
29
302. Supported tables
31-------------------
32
33 The following GX tables are currently supported.
34
35 bsln
36 feat
37 just
38 kern(*)
39 lcar
40 mort
41 morx
42 opbd
43 prop
44 trak
45
46 The following GX tables are currently unsupported.
47
48 cvar
49 fdsc
50 fmtx
51 fvar
52 gvar
53 Zapf
54
55 The following GX tables won't be supported.
56
57 acnt(**)
58 hsty(***)
59
60 The following undocumented tables in TrueType fonts designed for Apple
61 platform aren't handled either.
62
63 addg
64 CVTM
65 TPNM
66 umif
67
68
69 *) The `kern' validator handles both the classic and the new kern
70 formats; the former is supported on both Microsoft and Apple
71 platforms, while the latter is supported on Apple platforms.
72
73 **) `acnt' tables are not supported by currently available Apple font
74 tools.
75
76 ***) There is one more Apple extension, `hsty', but it is for
77 Newton-OS, not GX (Newton-OS is a platform by Apple, but it can
78 use sfnt- housed bitmap fonts only). Therefore, it should be
79 excluded from `Apple platform' in the context of TrueType.
80 gxvalid ignores it as Apple font tools do so.
81
82
83 We have checked 183 fonts bundled with MacOS 9.1, MacOS 9.2, MacOS
84 10.0, MacOS X 10.1, MSIE for MacOS, and AppleWorks 6.0. In addition,
85 we have checked 67 Dynalab fonts (designed for MacOS) and 189 Ricoh
86 fonts (designed for Windows and MacOS dual platforms). The number of
87 fonts including TrueType GX tables are as follows.
88
89 bsln: 76
90 feat: 191
91 just: 84
92 kern: 59
93 lcar: 4
94 mort: 326
95 morx: 19
96 opbd: 4
97 prop: 114
98 trak: 16
99
100 Dynalab and Ricoh fonts don't have GX tables except of `feat' and
101 `mort'.
102
103
1043. Fundamental validation limitations
105-------------------------------------
106
107 TrueType GX provides layout information to libraries for font
108 rasterizers and text layout. gxvalid can check whether the layout
109 data in a font is conformant to the TrueType GX format specified by
110 Apple. But gxvalid cannot check a how QuickDraw GX/AAT renderer uses
111 the stored information.
112
113 3-1. Validation of State Machine activity
114 -----------------------------------------
115
116 QuickDraw GX/AAT uses a `State Machine' to provide `stateful' layout
117 features, and TrueType GX stores the state transition diagram of
118 this `State Machine' in a `StateTable' data structure. While the
119 State Machine receives a series of glyph IDs, the State Machine
120 starts with `start of text' state, walks around various states and
121 generates various layout information to the renderer, and finally
122 reaches the `end of text' state.
123
124 gxvalid can check essential errors like:
125
126 - possibility of state transitions to undefined states
127 - existence of glyph IDs that the State Machine doesn't know how
128 to handle
129 - the State Machine cannot compute the layout information from
130 given diagram
131
132 These errors can be checked within finite steps, and without the
133 State Machine itself, because these are `expression' errors of state
134 transition diagram.
135
136 There is no limitation about how long the State Machine walks
137 around, so validation of the algorithm in the state transition
138 diagram requires infinite steps, even if we had a State Machine in
139 gxvalid. Therefore, the following errors and problems cannot be
140 checked.
141
142 - existence of states which the State Machine never transits to
143 - the possibility that the State Machine never reaches `end of
144 text'
145 - the possibility of stack underflow/overflow in the State Machine
146 (in ligature and contextual glyph substitutions, the State
147 Machine can store 16 glyphs onto its stack)
148
149 In addition, gxvalid doesn't check `temporary glyph IDs' used in the
150 chained State Machines (in `mort' and `morx' tables). If a layout
151 feature is implemented by a single State Machine, a glyph ID
152 converted by the State Machine is passed to the glyph renderer, thus
153 it should not point to an undefined glyph ID. But if a layout
154 feature is implemented by chained State Machines, a component State
155 Machine (if it is not the final one) is permitted to generate
156 undefined glyph IDs for temporary use, because it is handled by next
157 component State Machine and not by the glyph renderer. To validate
158 such temporary glyph IDs, gxvalid must stack all undefined glyph IDs
159 which can occur in the output of the previous State Machine and
160 search them in the `ClassTable' structure of the current State
161 Machine. It is too complex to list all possible glyph IDs from the
162 StateTable, especially from a ligature substitution table.
163
164 3-2. Validation of relationship between multiple layout features
165 ----------------------------------------------------------------
166
167 gxvalid does not validate the relationship between multiple layout
168 features at all.
169
170 If multiple layout features are defined in TrueType GX tables,
171 possible interactions, overrides, and conflicts between layout
172 features are implicitly given in the font too. For example, there
173 are several predefined spacing control features:
174
175 - Text Spacing (Proportional/Monospace/Half-width/Normal)
176 - Number Spacing (Monospaced-numbers/Proportional-numbers)
177 - Kana Spacing (Full-width/Proportional)
178 - Ideographic Spacing (Full-width/Proportional)
179 - CJK Roman Spacing (Half-width/Proportional/Default-roman
180 /Full-width-roman/Proportional)
181
182 If all layout features are independently managed, we can activate
183 inconsistent typographic rules like `Text Spacing=Monospace' and
184 `Ideographic Spacing=Proportional' at the same time.
185
186 The combinations of layout features is managed by a 32bit integer
187 (one bit each for selector setting), so we can define relationships
188 between up to 32 features, theoretically. But if one feature
189 setting affects another feature setting, we need typographic
190 priority rules to validate the relationship. Unfortunately, the
191 TrueType GX format specification does not give such information even
192 for predefined features.
193
194
1954. Permissive error handling of broken GX tables
196------------------------------------------------
197
198 When Apple's font rendering system finds an inconsistency, like a
199 specification violation or an unspecified value in a TrueType GX
200 table, it does not always return error. In most cases, the rendering
201 engine silently ignores such wrong values or even whole tables. In
202 fact, MacOS is shipped with fonts including broken GX/AAT tables, but
203 no harmful effects due to `officially broken' fonts are observed by
204 end-users.
205
206 gxvalid is designed to continue the validation process as long as
207 possible. When gxvalid find wrong values, gxvalid warns it at least,
208 and takes a fallback procedure if possible. The fallback procedure
209 depends on the debug level.
210
211 We used the following three tools to investigate Apple's error handling.
212
213 - FontValidator (for MacOS 8.5 - 9.2) resource fork font
214 - ftxvalidator (for MacOS X 10.1 -) dfont or naked-sfnt
215 - ftxdumperfuser (for MacOS X 10.1 -) dfont or naked-sfnt
216
217 However, all tests were done on a PowerPC based Macintosh; at present,
218 we have not checked those tools on a m68k-based Macintosh.
219
220 In total, we checked 183 fonts bundled to MacOS 9.1, MacOS 9.2, MacOS
221 10.0, MacOS X 10.1, MSIE for MacOS, and AppleWorks 6.0. These fonts
222 are distributed officially, but many broken GX/AAT tables were found
223 by Apple's font tools. In the following, we list typical violation of
224 the GX specification, in fonts officially distributed with those Apple
225 systems.
226
227 4-1. broken BinSrchHeader (19/183)
228 ----------------------------------
229
230 `BinSrchHeader' is a header of a data array for m68k platforms to
231 access memory efficiently. Although there are only two independent
232 parameters for real (`unitSize' and `nUnits'), BinSrchHeader has
233 three additional parameters which can be calculated from `unitSize'
234 and `nUnits', for fast setup. Apple font tools ignore them
235 silently, so gxvalid warns if it finds and inconsistency, and always
236 continues validation. The additional parameters are ignored
237 regardless of the consistency.
238
239 19 fonts include such inconsistencies; all breaks are in the
240 BinSrchHeader structure of the `kern' table.
241
242 4-2. too-short LookupTable (5/183)
243 ----------------------------------
244
245 LookupTable format 0 is a simple array to get a value from a given
246 GID (glyph ID); the index of this array is a GID too. Therefore,
247 the length of the array is expected to be same as the maximum GID
248 value defined in the `maxp' table, but there are some fonts whose
249 LookupTable format 0 is too short to cover all GIDs. FontValidator
250 ignores this error silently, ftxvalidator and ftxdumperfuser both
251 warn and continue. Similar problems are found in format 3 subtables
252 of `kern'. gxvalid warns always and abort if the validation level
253 is set to FT_VALIDATE_PARANOID.
254
255 5 fonts include too-short kern format 0 subtables.
256 1 font includes too-short kern format 3 subtable.
257
258 4-3. broken LookupTable format 2 (1/183)
259 ----------------------------------------
260
261 LookupTable format 2, subformat 4 covers the GID space by a
262 collection of segments which are specified by `firstGlyph' and
263 `lastGlyph'. Some fonts store `firstGlyph' and `lastGlyph' in
264 reverse order, so the segment specification is broken. Apple font
265 tools ignore this error silently; a broken segment is ignored as if
266 it did not exist. gxvalid warns and normalize the segment at
267 FT_VALIDATE_DEFAULT, or ignore the segment at FT_VALIDATE_TIGHT, or
268 abort at FT_VALIDATE_PARANOID.
269
270 1 font includes broken LookupTable format 2, in the `just' table.
271
272 *) It seems that all fonts manufactured by ITC for AppleWorks have
273 this error.
274
275 4-4. bad bracketing in glyph property (14/183)
276 ----------------------------------------------
277
278 GX/AAT defines a `bracketing' property of the glyphs in the `prop'
279 table, to control layout features of strings enclosed inside and
280 outside of brackets. Some fonts give inappropriate bracket
281 properties to glyphs. Apple font tools warn about this error;
282 gxvalid warns too and aborts at FT_VALIDATE_PARANOID.
283
284 14 fonts include wrong bracket properties.
285
286
287 4-5. invalid feature number (117/183)
288 -------------------------------------
289
290 The GX/AAT extension can include 255 different layout features,
291 but popular layout features are predefined (see
292 https://developer.apple.com/fonts/TrueType-Reference-Manual/RM09/AppendixF.html).
293 Some fonts include feature numbers which are incompatible with the
294 predefined feature registry.
295
296 In our survey, there are 140 fonts including `feat' table.
297
298 a) 67 fonts use a feature number which should not be used.
299 b) 117 fonts set the wrong feature range (nSetting). This is mostly
300 found in the `mort' and `morx' tables.
301
302 Apple font tools give no warning, although they cannot recognize
303 what the feature is. At FT_VALIDATE_DEFAULT, gxvalid warns but
304 continues in both cases (a, b). At FT_VALIDATE_TIGHT, gxvalid warns
305 and aborts for (a), but continues for (b). At FT_VALIDATE_PARANOID,
306 gxvalid warns and aborts in both cases (a, b).
307
308 4-6. invalid prop version (10/183)
309 ----------------------------------
310
311 As most TrueType GX tables, the `prop' table must start with a 32bit
312 version identifier: 0x00010000, 0x00020000 or 0x00030000. But some
313 fonts store nonsense binary data instead. When Apple font tools
314 find them, they abort the processing immediately, and the data which
315 follows is unhandled. gxvalid does the same.
316
317 10 fonts include broken `prop' version.
318
319 All of these fonts are classic TrueType fonts for the Japanese
320 script, manufactured by Apple.
321
322 4-7. unknown resource name (2/183)
323 ------------------------------------
324
325 NOTE: THIS IS NOT A TRUETYPE GX ERROR.
326
327 If a TrueType font is stored in the resource fork or in dfont
328 format, the data must be tagged as `sfnt' in the resource fork index
329 to invoke TrueType font handler for the data. But the TrueType font
330 data in `Keyboard.dfont' is tagged as `kbd', and that in
331 `LastResort.dfont' is tagged as `lst'. Apple font tools can detect
332 that the data is in TrueType format and successfully validate them.
333 Maybe this is possible because they are known to be dfont. The
334 current implementation of the resource fork driver of FreeType
335 cannot do that, thus gxvalid cannot validate them.
336
337 2 fonts use an unknown tag for the TrueType font resource.
338
3395. `kern' table issues
340----------------------
341
342 In common terminology of TrueType, `kern' is classified as a basic and
343 platform-independent table. But there are Apple extensions of `kern',
344 and there is an extension which requires a GX state machine for
345 contextual kerning. Therefore, gxvalid includes a special validator
346 for `kern' tables. Unfortunately, there is no exact algorithm to
347 check Apple's extension, so gxvalid includes a heuristic algorithm to
348 find the proper validation routines for all possible data formats,
349 including the data format for Microsoft. By calling
350 classic_kern_validate() instead of gxv_validate(), you can specify the
351 `kern' format explicitly. However, current FreeType2 uses Microsoft
352 `kern' format only, others are ignored (and should be handled in a
353 library one level higher than FreeType).
354
355 5-1. History
356 ------------
357
358 The original 16bit version of `kern' was designed by Apple in the
359 pre-GX era, and it was also approved by Microsoft. Afterwards,
360 Apple designed a new 32bit version of the `kern' table. According
361 to the documentation, the difference between the 16bit and 32bit
362 version is only the size of variables in the `kern' header. In the
363 following, we call the original 16bit version as `classic', and
364 32bit version as `new'.
365
366 5-2. Versions and dialects which should be differentiated
367 ---------------------------------------------------------
368
369 The `kern' table consists of a table header and several subtables.
370 The version number which identifies a `classic' or a `new' version
371 is explicitly written in the table header, but there are
372 undocumented differences between Microsoft's and Apple's formats.
373 It is called a `dialect' in the following. There are three cases
374 which should be handled: the new Apple-dialect, the classic
375 Apple-dialect, and the classic Microsoft-dialect. An analysis of
376 the formats and the auto detection algorithm of gxvalid is described
377 in the following.
378
379 5-2-1. Version detection: classic and new kern
380 ----------------------------------------------
381
382 According to Apple TrueType specification, there are only two
383 differences between the classic and the new:
384
385 - The `kern' table header starts with the version number.
386 The classic version starts with 0x0000 (16bit),
387 the new version starts with 0x00010000 (32bit).
388
389 - In the `kern' table header, the number of subtables follows
390 the version number.
391 In the classic version, it is stored as a 16bit value.
392 In the new version, it is stored as a 32bit value.
393
394 From Apple font tool's output (DumpKERN is also tested in addition
395 to the three Apple font tools in above), there is another
396 undocumented difference. In the new version, the subtable header
397 includes a 16bit variable named `tupleIndex' which does not exist
398 in the classic version.
399
400 The new version can store all subtable formats (0, 1, 2, and 3),
401 but the Apple TrueType specification does not mention the subtable
402 formats available in the classic version.
403
404 5-2-2. Available subtable formats in classic version
405 ----------------------------------------------------
406
407 Although the Apple TrueType specification recommends to use the
408 classic version in the case if the font is designed for both the
409 Apple and Microsoft platforms, it does not document the available
410 subtable formats in the classic version.
411
412 According to the Microsoft TrueType specification, the subtable
413 format assured for Windows and OS/2 support is only subtable
414 format 0. The Microsoft TrueType specification also describes
415 subtable format 2, but does not mention which platforms support
416 it. Subtable formats 1, 3, and higher are documented as reserved
417 for future use. Therefore, the classic version can store subtable
418 formats 0 and 2, at least. `ttfdump.exe', a font tool provided by
419 Microsoft, ignores the subtable format written in the subtable
420 header, and parses the table as if all subtables are in format 0.
421
422 `kern' subtable format 1 uses a StateTable, so it cannot be
423 utilized without a GX State Machine. Therefore, it is reasonable
424 to assume that format 1 (and 3) were introduced after Apple had
425 introduced GX and moved to the new 32bit version.
426
427 5-2-3. Apple and Microsoft dialects
428 -----------------------------------
429
430 The `kern' subtable has a 16bit `coverage' field to describe
431 kerning attributes, but bit interpretations by Apple and Microsoft
432 are different: For example, Apple uses bits 0-7 to identify the
433 subtable, while Microsoft uses bits 8-15.
434
435 In addition, due to the output of DumpKERN and FontValidator,
436 Apple's bit interpretations of coverage in classic and new version
437 are incompatible also. In summary, there are three dialects:
438 classic Apple dialect, classic Microsoft dialect, and new Apple
439 dialect. The classic Microsoft dialect and the new Apple dialect
440 are documented by each vendors' TrueType font specification, but
441 the documentation for classic Apple dialect is not available.
442
443 For example, in the new Apple dialect, bit 15 is documented as
444 `set to 1 if the kerning is vertical'. On the other hand, in
445 classic Microsoft dialect, bit 1 is documented as `set to 1 if the
446 kerning is horizontal'. From the outputs of DumpKERN and
447 FontValidator, classic Apple dialect recognizes 15 as `set to 1
448 when the kerning is horizontal'. From the results of similar
449 experiments, classic Apple dialect seems to be the Endian reverse
450 of the classic Microsoft dialect.
451
452 As a conclusion it must be noted that no font tool can identify
453 classic Apple dialect or classic Microsoft dialect automatically.
454
455 5-2-4. gxvalid auto dialect detection algorithm
456 -----------------------------------------------
457
458 The first 16 bits of the `kern' table are enough to identify the
459 version:
460
461 - if the first 16 bits are 0x0000, the `kern' table is in
462 classic Apple dialect or classic Microsoft dialect
463 - if the first 16 bits are 0x0001, and next 16 bits are 0x0000,
464 the kern table is in new Apple dialect.
465
466 If the `kern' table is a classic one, the 16bit `coverage' field
467 is checked next. Firstly, the coverage bits are decoded for the
468 classic Apple dialect using the following bit masks (this is based
469 on DumpKERN output):
470
471 0x8000: 1=horizontal, 0=vertical
472 0x4000: not used
473 0x2000: 1=cross-stream, 0=normal
474 0x1FF0: reserved
475 0x000F: subtable format
476
477 If any of reserved bits are set or the subtable bits is
478 interpreted as format 1 or 3, we take it as `impossible in classic
479 Apple dialect' and retry, using the classic Microsoft dialect.
480
481 The most popular coverage in new Apple-dialect: 0x8000,
482 The most popular coverage in classic Apple-dialect: 0x0000,
483 The most popular coverage in classic Microsoft dialect: 0x0001.
484
485 5-3. Tested fonts
486 -----------------
487
488 We checked 59 fonts bundled with MacOS and 38 fonts bundled with
489 Windows, where all font include a `kern' table.
490
491 - fonts bundled with MacOS
492 * new Apple dialect
493 format 0: 18
494 format 2: 1
495 format 3: 1
496 * classic Apple dialect
497 format 0: 14
498 * classic Microsoft dialect
499 format 0: 15
500
501 - fonts bundled with Windows
502 * classic Microsoft dialect
503 format 0: 38
504
505 It looks strange that classic Microsoft-dialect fonts are bundled to
506 MacOS: they come from MSIE for MacOS, except of MarkerFelt.dfont.
507
508
509 ACKNOWLEDGEMENT
510 ---------------
511
512 Some parts of gxvalid are derived from both the `gxlayout' module and
513 the `otvalid' module. Development of gxlayout was supported by the
514 Information-technology Promotion Agency(IPA), Japan.
515
516 The detailed analysis of undefined glyph ID utilization in `mort' and
517 `morx' tables is provided by George Williams.
518
519------------------------------------------------------------------------
520
521Copyright (C) 2004-2020 by
522suzuki toshiya, Masatake YAMATO, Red hat K.K.,
523David Turner, Robert Wilhelm, and Werner Lemberg.
524
525This file is part of the FreeType project, and may only be used,
526modified, and distributed under the terms of the FreeType project
527license, LICENSE.TXT. By continuing to use, modify, or distribute this
528file you indicate that you have read the license and understand and
529accept it fully.
530
531
532--- end of README ---
533