summaryrefslogtreecommitdiffstats
path: root/TODO
blob: 9061d1500f2b3099ea3b86475a1822ba6558e4e9 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
************************************************************************
* Official mandoc TODO.
* $Id$
************************************************************************

************************************************************************
* crashes
************************************************************************

- The abort() in bufcat(), html.c, can be triggered via buffmt_includes()
  by running -Thtml -Oincludes on a file containing a long .In argument.
  Fixing this will probably require reworking the whole bufcat() concept.

************************************************************************
* missing features
************************************************************************

--- missing roff features ----------------------------------------------

- .ad (adjust margins)
  .ad l -- adjust left margin only (flush left)
  .ad r -- adjust right margin only (flush right)
  .ad c -- center text on line
  .ad b -- adjust both margins (alias: .ad n)
  .na   -- temporarily disable adjustment without changing the mode
  .ad   -- re-enable adjustment without changing the mode
  Adjustment mode is ignored while in no-fill mode (.nf).

- .fc (field control)
  found by naddy@ in xloadimage(1)
  
- .nr third argument (auto-increment step size, requires \n+)
  found by bentley@ in sbcl(1)  Mon, 9 Dec 2013 18:36:57 -0700

- .ns (no-space mode) occurs in xine-config(1)
  reported by brad@  Sat, 15 Jan 2011 15:45:23 -0500

- .ta (tab settings) occurs in ircbug(1) and probably gnats(1)
  reported by brad@  Sat, 15 Jan 2011 15:50:51 -0500
  also Tcl_NewStringObj(3) via wiz@  Wed, 5 Mar 2014 22:27:43 +0100

- .ti (temporary indent)
  found by naddy@ in xloadimage(1)
  found by bentley@ in nmh(1)  Mon, 23 Apr 2012 13:38:28 -0600

- .while and .shift 
  found by jca@ in ratpoison(1)  Sun, 30 Jun 2013 12:01:09 +0200

- \c (interrupted text) should prevent the line break
  even inside .Bd literal; that occurs in chat(8)
  also found in cclive(1) - DocBook output

- \h horizontal move
  found in cclive(1) DocBook output
  Anthony J. Bentley on discuss@  Sat, 21 Sep 2013 22:29:34 -0600

- \n+ and \n- numerical register increment and decrement
  found by bentley@ in sbcl(1)  Mon, 9 Dec 2013 18:36:57 -0700

- \w'' width measurements
  would not be very useful without an expression parser, see below
  needed for Tcl_NewStringObj(3) via wiz@  Wed, 5 Mar 2014 22:27:43 +0100

- using undefined strings or macros defines them to be empty
  wl@  Mon, 14 Nov 2011 14:37:01 +0000

--- missing mdoc features ----------------------------------------------

- fix bad block nesting involving multiple identical explicit blocks
  see the OpenBSD mdoc_macro.c 1.47 commit message

- .Bl -column .Xo support is missing
  ultimate goal:
  restore .Xr and .Dv to
  lib/libc/compat-43/sigvec.3
  lib/libc/gen/signal.3
  lib/libc/sys/sigaction.2

- edge case: decide how to deal with blk_full bad nesting, e.g.
  .Sh .Nm .Bk .Nm .Ek .Sh found by jmc@ in ssh-keygen(1)
  from jmc@  Wed, 14 Jul 2010 18:10:32 +0100

- \\ is now implemented correctly
  * when defining strings and macros using .ds and .de
  * when parsing roff(7) and man(7) macro arguments
  It does not yet work in mdoc(7) macro arguments
  because libmdoc does not yet use mandoc_getarg().
  Also check what happens in plain text, it must be identical to \e.

- .Bd -centered implies -filled, not -unfilled, which is not
  easy to implement; it requires code similar to .ce, which
  we don't have either.
  Besides, groff has bug causing text right *before* .Bd -centered
  to be centered as well.

- .Bd -filled should not be the same as .Bd -ragged, but align both
  the left and right margin.  In groff, it is implemented in terms
  of .ad b, which we don't have either.  Found in cksum(1).

- implement blank `Bl -column', such as
  .Bl -column
  .It foo Ta bar
  .El

- explicitly disallow nested `Bl -column', which would clobber internal
  flags defined for struct mdoc_macro

- In .Bl -column .It, the end of the line probably has to be regarded
  as an implicit .Ta, if there could be one, see the following mildly
  ugly code from login.conf(5):
    .Bl -column minpasswordlen program xetcxmotd
    .It path Ta path Ta value of Dv _PATH_DEFPATH
    .br
    Default search path.
  reported by Michal Mazurek <akfaew at jasminek dot net>
  via jmc@ Thu, 7 Apr 2011 16:00:53 +0059

- inside `.Bl -column' phrases, punctuation is handled like normal
  text, e.g. `.Bl -column .It Fl x . Ta ...' should give "-x -."

- inside `.Bl -column' phrases, TERMP_IGNDELIM handling by `Pf'
  is not safe, e.g. `.Bl -column .It Pf a b .' gives "ab."
  but should give "ab ."

- set a meaningful default if no `Bl' list type is assigned

- have a blank `It' head for `Bl -tag' not puke

- check whether it is correct that `D1' uses INDENT+1;
  does it need its own constant?

- prohibit `Nm' from having non-text HEAD children
  (e.g., NetBSD mDNSShared/dns-sd.1)
  (mdoc_html.c and mdoc_term.c `Nm' handlers can be slightly simplified)

- support translated section names
  e.g. x11/scrotwm scrotwm_es.1:21:2: error: NAME section must be first
  that one uses NOMBRE because it is spanish...
  deraadt tends to think that section-dependent macro behaviour
  is a bad idea in the first place, so this may be irrelevant

- When there is free text in the SYNOPSIS and that free text contains
  the .Nm macro, groff somehow understands to treat the .Nm as an in-line
  macro, while mandoc treats it as a block macro and breaks the line.
  No idea how the logic for distinguishing in-line and block instances
  should be, needs investigation.
  uqs@  Thu, 2 Jun 2011 11:03:51 +0200
  uqs@  Thu, 2 Jun 2011 11:33:35 +0200

--- missing man features -----------------------------------------------

- -T[x]html doesn't stipulate non-collapsing spaces in literal mode

--- missing tbl features -----------------------------------------------

- look at the POSIX manuals in the books/man-pages-posix port,
  they use some unsupported tbl(7) features.

- investigate tbl(1) errors in sox(1)
  see also naddy@  Sat, 16 Oct 2010 23:51:57 +0200

- allow standalone `.' to be interpreted as an end-of-layout
  delimiter instead of being thrown away as a no-op roff line
  reported by Yuri Pankov, Wed 18 May 2011 11:34:59 CEST

--- missing misc features ----------------------------------------------

- italic correction (\/) in PostScript mode
  Werner LEMBERG on groff at gnu dot org  Sun, 10 Nov 2013 12:47:46

- When makewhatis(8) encounters a FATAL parse error,
  it silently treats the file as formatted, which makes no sense
  at all for paths like man1/foo.1 - and which also contradicts
  what the manual says at the end of the description.
  The end result will be ENOENT for file names returned
  by mansearch() in manpage.file.

- makewhatis(8) for preformatted pages:
  parse the section number from the header line
  and compare to the section number from the directory name

- Does makewhatis(8) detect missing NAME sections, missing names,
  and missing descriptions in all the file formats?

- clean up escape sequence handling, creating three classes:
  (1) fully implemented, or parsed and ignored without loss of content
  (2) unimplemented, potentially causing loss of content
      or serious mangling of formatting (e.g. \n) -> ERROR
      see textproc/mgdiff(1) for nice examples
  (3) undefined, just output the character -> perhaps WARNING

- kettenis wants base roff, ms, and me  Fri, 1 Jan 2010 22:13:15 +0100 (CET)

--- compatibility checks -----------------------------------------------

- is .Bk implemented correctly in modern groff?
  sobrado@  Tue, 19 Apr 2011 22:12:55 +0200

- compare output to Heirloom roff, Solaris roff, and
  http://repo.or.cz/w/neatroff.git  http://litcave.rudi.ir/

- look at pages generated from reStructeredText, e.g. devel/mercurial hg(1)
  These are a weird mixture of man(7) and custom autogenerated low-level
  roff stuff.  Figure out to what extent we can cope.
  For details, see http://docutils.sourceforge.net/rst.html
  noted by stsp@  Sat, 24 Apr 2010 09:17:55 +0200
  reminded by nicm@  Mon, 3 May 2010 09:52:41 +0100

- look at pages generated from ronn(1) github.com/rtomayko/ronn
  (based on markdown)

- look at pages generated from Texinfo source by yat2m, e.g. security/gnupg
  First impression is not that bad.

- look at pages generated by pandoc; see
  https://github.com/jgm/pandoc/blob/master/src/Text/Pandoc/Writers/Man.hs
  porting planned by kili@  Thu, 19 Jun 2014 19:46:28 +0200

- check compatibility with Plan9:
  http://swtch.com/usr/local/plan9/tmac/tmac.an
  http://swtch.com/plan9port/man/man7/man.html
  "Anthony J. Bentley" <anthonyjbentley@gmail.com> 28 Dec 2010 21:58:40 -0700

- check compatibility with the man(7) formatter
  https://raw.githubusercontent.com/rofl0r/hardcore-utils/master/man.c

************************************************************************
* formatting issues: ugly output
************************************************************************

- a column list with blank `Ta' cells triggers a spurrious
  start-with-whitespace printing of a newline

- In .Bl -column,
  .It Em Authentication<tab>Key Length
  ought to render "Key Length" with emphasis, too,
  see OpenBSD iked.conf(5).
  reported again Nicolas Joly via wiz@ Wed, 12 Oct 2011 00:20:00 +0200

- empty phrases in .Bl column produce too few blanks
  try e.g. .Bl -column It Ta Ta
  reported by millert Fri, 02 Apr 2010 16:13:46 -0400

- .%T can have trailing punctuation.  Currently, it puts the trailing
  punctuation into a trailing MDOC_TEXT element inside its own scope.
  That element should rather be outside its scope, such that the
  punctuation does not get underlines.  This is not trivial to
  implement because .%T then needs some features of in_line_eoln() -
  slurp all arguments into one single text element - and one feature
  of in_line() - put trailing punctuation out of scope.
  Found in mount_nfs(8) and exports(5), search for "Appendix".

- Trailing punctuation after .%T triggers EOS spacing, at least
  outside .Rs (eek!).  Simply setting ARGSFL_DELIM for .%T is not
  the right solution, it sends mandoc into an endless loop.
  reported by Nicolas Joly  Sat, 17 Nov 2012 11:49:54 +0100

- global variables in the SYNOPSIS of section 3 pages
  .Vt vs .Vt/.Va vs .Ft/.Va vs .Ft/.Fa ...
  from kristaps@  Tue, 08 Jun 2010 11:13:32 +0200

- in enclosures, mandoc sometimes fancies a bogus end of sentence
  reminded by jmc@  Thu, 23 Sep 2010 18:13:39 +0059

- formatting /usr/local/man/man1/latex2man.1 with groff and mandoc
  reveals lots of bugs both in groff and mandoc...
  reported by bentley@  Wed, 22 May 2013 23:49:30 -0600

--- PDF issues ---------------------------------------------------------

- PDF output doesn't use a monospaced font for .Bd -literal
  Example: "mandoc -Tpdf afterboot.8 > output.pdf && pdfviewer output.pdf".
  Search the text "Routing tables".
  Also check what PostScript mode does when fixing this.
  reported by juanfra@ Wed, 04 Jun 2014 21:44:58 +0200

--- HTML issues --------------------------------------------------------

- <dl><dt><dd> formatting is ugly
  hints are easy to find on the web, e.g.
  http://stackoverflow.com/questions/1713048/
  see also matthew@  Fri, 18 Jul 2014 19:25:12 -0700

- check https://github.com/trentm/mdocml

************************************************************************
* formatting issues: gratuitous differences
************************************************************************

- .Rv (and probably .Ex) print different text if an `Nm' has been named
  or not (run a manual without `Nm blah' to see this).  I'm not sure
  that this exists in the wild, but it's still an error.

- In .Bl -bullet, the groff bullet is "+\b+\bo\bo", the mandoc bullet
  is just "o\bo".
  see for example OpenBSD ksh(1)

- In .Bl -enum -width 0n, groff continues one the same line after
  the number, mandoc breaks the line.
  mail to kristaps@  Mon, 20 Jul 2009 02:21:39 +0200

- .Pp between two .It in .Bl -column should produce one,
  not two blank lines, see e.g. login.conf(5).
  reported by jmc@  Sun, 17 Apr 2011 14:04:58 +0059
  reported again by sthen@  Wed, 18 Jan 2012 02:09:39 +0000 (UTC)

- If the *first* line after .It is .Pp, break the line right after
  the tag, do not pad with space characters before breaking.
  See the description of the a, c, and i commands in sed(1).

- If the first line after .It is .D1, do not assert a blank line
  in between, see for example tmux(1).
  reported by nicm@  13 Jan 2011 00:18:57 +0000

- Trailing punctuation after .It should trigger EOS spacing.
  reported by Nicolas Joly  Sat, 17 Nov 2012 11:49:54 +0100
  Probably, this should be fixed somewhere in termp_it_pre(), not sure.

- .Nx 1.0a
  should be "NetBSD 1.0A", not "NetBSD 1.0a",
  see OpenBSD ccdconfig(8).

- In .Bl -tag, if a tag exceeds the right margin and must be continued
  on the next line, it must be indented by -width, not width+1;
  see "rule block|pass" in OpenBSD ifconfig(8).

- When the -width string contains macros, the macros must be rendered
  before measuring the width, for example
    .Bl -tag -width ".Dv message"
  in magic(5), located in src/usr.bin/file, is the same
  as -width 7n, not -width 11n.
  The same applies to .Bl -column column widths;
  reported again by Nicolas Joly Thu, 1 Mar 2012 13:41:26 +0100 via wiz@ 5 Mar
  reported again by Franco Fichtner Fri, 27 Sep 2013 21:02:28 +0200
  An easy partial fix would be to just skip the first word if it starts
  with a dot, including any following white space, when measuring.

- The \& zero-width character counts as output.
  That is, when it is alone on a line between two .Pp,
  we want three blank lines, not two as in mandoc.

- Header lines of excessive length:
  Port OpenBSD man_term.c rev. 1.25 to mdoc_term.c
  and document it in mdoc(7) and man(7) COMPATIBILITY
  found while talking to Chris Bennett

- trailing whitespace must be ignored even when followed by a font escape,
  see for example 
    makes
    \fBdig \fR
    operate in batch mode
  in dig(1).

************************************************************************
* warning issues
************************************************************************

- check that MANDOCERR_BADTAB is thrown in the right cases,
  i.e. when finding a literal tab character in fill mode,
  and possibly change the wording of the warning message
  to refer to fill mode, not literal mode
  See the mail from Werner LEMBERG on the groff list,
  Fri, 14 Feb 2014 18:54:42 +0100 (CET)

- warn about "new sentence, new line"

- mandoc_special does not really check the escape sequence,
  but just the overall format

- integrate mdoclint into mandoc ("end-of-line whitespace" thread)
  from jmc@  Mon, 13 Jul 2009 17:12:09 +0100
  from kristaps@  Mon, 13 Jul 2009 18:34:53 +0200
  from jmc@  Mon, 13 Jul 2009 17:45:37 +0059
  from kristaps@  Mon, 13 Jul 2009 19:02:03 +0200

- -Tlint parser errors and warnings to stdout
  to tech@mdocml, naddy@  Wed, 28 Sep 2011 11:21:46 +0200
  wait!  kristaps@  Sun, 02 Oct 2011 17:12:52 +0200

- for system errors, use errno/strerror/warn/err

************************************************************************
* documentation issues
************************************************************************

- mention hyphenation rules:
  breaking at letter-letter in text mode (not macro args)
  proper hyphenation is unimplemented

- talk about spacing around delimiters
  to jmc@, kristaps@  Sat, 23 Apr 2011 17:41:27 +0200

- mark macros as: page structure domain, manual domain, general text domain
  is this useful?

- mention /usr/share/misc/mdoc.template in mdoc(7)?

************************************************************************
* performance issues
************************************************************************

- Why are we using MAP_SHARED, not MAP_PRIVATE for mmap(2)?
  How does SQLITE_CONFIG_PAGECACHE actually work?  Document it!
  from kristaps@  Sat, 09 Aug 2014 13:51:36 +0200

Several areas can be cleaned up to make mandoc even faster.  These are 

- improve hashing mechanism for macros (quite important: performance)

- improve hashing mechanism for characters (not as important)

- the PDF file is HUGE: this can be reduced by using relative offsets

- instead of re-initialising the roff predefined-strings set before each
  parse, create a read-only version the first time and copy it 

************************************************************************
* structural issues
************************************************************************

- We use the input line number at several places to distinguish
  same-line from different-line input.  That plainly doesn't work
  with user-defined macros, leading to random breakage.

- Find better ways to prevent endless loops
  in roff(7) macro and string expansion.
 
- Finish cleanup of date handling.
  Decide which formats should be recognized where.
  Update both mdoc(7) and man(7) documentation.
  Triggered by  Tim van der Molen  Tue, 22 Feb 2011 20:30:45 +0100

- Consider creating some views that will make the database more
  readable from the sqlite3 shell.  Consider using them to
  abstract from the database structure, too.
  suggested by espie@  Sat, 19 Apr 2014 14:52:57 +0200