1 # SPDX-License-Identifier: GPL-2.0 1 # SPDX-License-Identifier: GPL-2.0
2 # Copyright 2019 Jonathan Corbet <corbet@lwn.ne 2 # Copyright 2019 Jonathan Corbet <corbet@lwn.net>
3 # 3 #
4 # Apply kernel-specific tweaks after the initi 4 # Apply kernel-specific tweaks after the initial document processing
5 # has been done. 5 # has been done.
6 # 6 #
7 from docutils import nodes 7 from docutils import nodes
8 import sphinx 8 import sphinx
9 from sphinx import addnodes 9 from sphinx import addnodes
10 from sphinx.errors import NoUri !! 10 if sphinx.version_info[0] < 2 or \
>> 11 sphinx.version_info[0] == 2 and sphinx.version_info[1] < 1:
>> 12 from sphinx.environment import NoUri
>> 13 else:
>> 14 from sphinx.errors import NoUri
11 import re 15 import re
12 from itertools import chain 16 from itertools import chain
13 17
14 # 18 #
15 # Python 2 lacks re.ASCII... 19 # Python 2 lacks re.ASCII...
16 # 20 #
17 try: 21 try:
18 ascii_p3 = re.ASCII 22 ascii_p3 = re.ASCII
19 except AttributeError: 23 except AttributeError:
20 ascii_p3 = 0 24 ascii_p3 = 0
21 25
22 # 26 #
23 # Regex nastiness. Of course. 27 # Regex nastiness. Of course.
24 # Try to identify "function()" that's not alre 28 # Try to identify "function()" that's not already marked up some
25 # other way. Sphinx doesn't like a lot of stu 29 # other way. Sphinx doesn't like a lot of stuff right after a
26 # :c:func: block (i.e. ":c:func:`mmap()`s" fla 30 # :c:func: block (i.e. ":c:func:`mmap()`s" flakes out), so the last
27 # bit tries to restrict matches to things that 31 # bit tries to restrict matches to things that won't create trouble.
28 # 32 #
29 RE_function = re.compile(r'\b(([a-zA-Z_]\w+)\( 33 RE_function = re.compile(r'\b(([a-zA-Z_]\w+)\(\))', flags=ascii_p3)
30 34
31 # 35 #
32 # Sphinx 2 uses the same :c:type role for stru 36 # Sphinx 2 uses the same :c:type role for struct, union, enum and typedef
33 # 37 #
34 RE_generic_type = re.compile(r'\b(struct|union 38 RE_generic_type = re.compile(r'\b(struct|union|enum|typedef)\s+([a-zA-Z_]\w+)',
35 flags=ascii_p3) 39 flags=ascii_p3)
36 40
37 # 41 #
38 # Sphinx 3 uses a different C role for each on 42 # Sphinx 3 uses a different C role for each one of struct, union, enum and
39 # typedef 43 # typedef
40 # 44 #
41 RE_struct = re.compile(r'\b(struct)\s+([a-zA-Z 45 RE_struct = re.compile(r'\b(struct)\s+([a-zA-Z_]\w+)', flags=ascii_p3)
42 RE_union = re.compile(r'\b(union)\s+([a-zA-Z_] 46 RE_union = re.compile(r'\b(union)\s+([a-zA-Z_]\w+)', flags=ascii_p3)
43 RE_enum = re.compile(r'\b(enum)\s+([a-zA-Z_]\w 47 RE_enum = re.compile(r'\b(enum)\s+([a-zA-Z_]\w+)', flags=ascii_p3)
44 RE_typedef = re.compile(r'\b(typedef)\s+([a-zA 48 RE_typedef = re.compile(r'\b(typedef)\s+([a-zA-Z_]\w+)', flags=ascii_p3)
45 49
46 # 50 #
47 # Detects a reference to a documentation page 51 # Detects a reference to a documentation page of the form Documentation/... with
48 # an optional extension 52 # an optional extension
49 # 53 #
50 RE_doc = re.compile(r'(\bDocumentation/)?((\.\ 54 RE_doc = re.compile(r'(\bDocumentation/)?((\.\./)*[\w\-/]+)\.(rst|txt)')
51 55
52 RE_namespace = re.compile(r'^\s*..\s*c:namespa 56 RE_namespace = re.compile(r'^\s*..\s*c:namespace::\s*(\S+)\s*$')
53 57
54 # 58 #
55 # Reserved C words that we should skip when cr 59 # Reserved C words that we should skip when cross-referencing
56 # 60 #
57 Skipnames = [ 'for', 'if', 'register', 'sizeof 61 Skipnames = [ 'for', 'if', 'register', 'sizeof', 'struct', 'unsigned' ]
58 62
59 63
60 # 64 #
61 # Many places in the docs refer to common syst 65 # Many places in the docs refer to common system calls. It is
62 # pointless to try to cross-reference them and 66 # pointless to try to cross-reference them and, as has been known
63 # to happen, somebody defining a function by t 67 # to happen, somebody defining a function by these names can lead
64 # to the creation of incorrect and confusing c 68 # to the creation of incorrect and confusing cross references. So
65 # just don't even try with these names. 69 # just don't even try with these names.
66 # 70 #
67 Skipfuncs = [ 'open', 'close', 'read', 'write' 71 Skipfuncs = [ 'open', 'close', 'read', 'write', 'fcntl', 'mmap',
68 'select', 'poll', 'fork', 'execv 72 'select', 'poll', 'fork', 'execve', 'clone', 'ioctl',
69 'socket' ] 73 'socket' ]
70 74
71 c_namespace = '' 75 c_namespace = ''
72 76
73 # <<
74 # Detect references to commits. <<
75 # <<
76 RE_git = re.compile(r'commit\s+(?P<rev>[0-9a-f <<
77 flags=re.IGNORECASE | re.DOTALL) <<
78 <<
79 def markup_refs(docname, app, node): 77 def markup_refs(docname, app, node):
80 t = node.astext() 78 t = node.astext()
81 done = 0 79 done = 0
82 repl = [ ] 80 repl = [ ]
83 # 81 #
84 # Associate each regex with the function t 82 # Associate each regex with the function that will markup its matches
85 # 83 #
86 markup_func_sphinx2 = {RE_doc: markup_doc_ 84 markup_func_sphinx2 = {RE_doc: markup_doc_ref,
87 RE_function: markup 85 RE_function: markup_c_ref,
88 RE_generic_type: ma 86 RE_generic_type: markup_c_ref}
89 87
90 markup_func_sphinx3 = {RE_doc: markup_doc_ 88 markup_func_sphinx3 = {RE_doc: markup_doc_ref,
91 RE_function: markup 89 RE_function: markup_func_ref_sphinx3,
92 RE_struct: markup_c 90 RE_struct: markup_c_ref,
93 RE_union: markup_c_ 91 RE_union: markup_c_ref,
94 RE_enum: markup_c_r 92 RE_enum: markup_c_ref,
95 RE_typedef: markup_ !! 93 RE_typedef: markup_c_ref}
96 RE_git: markup_git} <<
97 94
98 if sphinx.version_info[0] >= 3: 95 if sphinx.version_info[0] >= 3:
99 markup_func = markup_func_sphinx3 96 markup_func = markup_func_sphinx3
100 else: 97 else:
101 markup_func = markup_func_sphinx2 98 markup_func = markup_func_sphinx2
102 99
103 match_iterators = [regex.finditer(t) for r 100 match_iterators = [regex.finditer(t) for regex in markup_func]
104 # 101 #
105 # Sort all references by the starting posi 102 # Sort all references by the starting position in text
106 # 103 #
107 sorted_matches = sorted(chain(*match_itera 104 sorted_matches = sorted(chain(*match_iterators), key=lambda m: m.start())
108 for m in sorted_matches: 105 for m in sorted_matches:
109 # 106 #
110 # Include any text prior to match as a 107 # Include any text prior to match as a normal text node.
111 # 108 #
112 if m.start() > done: 109 if m.start() > done:
113 repl.append(nodes.Text(t[done:m.st 110 repl.append(nodes.Text(t[done:m.start()]))
114 111
115 # 112 #
116 # Call the function associated with th 113 # Call the function associated with the regex that matched this text and
117 # append its return to the text 114 # append its return to the text
118 # 115 #
119 repl.append(markup_func[m.re](docname, 116 repl.append(markup_func[m.re](docname, app, m))
120 117
121 done = m.end() 118 done = m.end()
122 if done < len(t): 119 if done < len(t):
123 repl.append(nodes.Text(t[done:])) 120 repl.append(nodes.Text(t[done:]))
124 return repl 121 return repl
125 122
126 # 123 #
127 # Keep track of cross-reference lookups that f 124 # Keep track of cross-reference lookups that failed so we don't have to
128 # do them again. 125 # do them again.
129 # 126 #
130 failed_lookups = { } 127 failed_lookups = { }
131 def failure_seen(target): 128 def failure_seen(target):
132 return (target) in failed_lookups 129 return (target) in failed_lookups
133 def note_failure(target): 130 def note_failure(target):
134 failed_lookups[target] = True 131 failed_lookups[target] = True
135 132
136 # 133 #
137 # In sphinx3 we can cross-reference to C macro 134 # In sphinx3 we can cross-reference to C macro and function, each one with its
138 # own C role, but both match the same regex, s 135 # own C role, but both match the same regex, so we try both.
139 # 136 #
140 def markup_func_ref_sphinx3(docname, app, matc 137 def markup_func_ref_sphinx3(docname, app, match):
141 cdom = app.env.domains['c'] 138 cdom = app.env.domains['c']
142 # 139 #
143 # Go through the dance of getting an xref 140 # Go through the dance of getting an xref out of the C domain
144 # 141 #
145 base_target = match.group(2) 142 base_target = match.group(2)
146 target_text = nodes.Text(match.group(0)) 143 target_text = nodes.Text(match.group(0))
147 xref = None 144 xref = None
148 possible_targets = [base_target] 145 possible_targets = [base_target]
149 # Check if this document has a namespace, 146 # Check if this document has a namespace, and if so, try
150 # cross-referencing inside it first. 147 # cross-referencing inside it first.
151 if c_namespace: 148 if c_namespace:
152 possible_targets.insert(0, c_namespace 149 possible_targets.insert(0, c_namespace + "." + base_target)
153 150
154 if base_target not in Skipnames: 151 if base_target not in Skipnames:
155 for target in possible_targets: 152 for target in possible_targets:
156 if (target not in Skipfuncs) and n 153 if (target not in Skipfuncs) and not failure_seen(target):
157 lit_text = nodes.literal(class 154 lit_text = nodes.literal(classes=['xref', 'c', 'c-func'])
158 lit_text += target_text 155 lit_text += target_text
159 pxref = addnodes.pending_xref( 156 pxref = addnodes.pending_xref('', refdomain = 'c',
160 157 reftype = 'function',
161 158 reftarget = target,
162 159 modname = None,
163 160 classname = None)
164 # 161 #
165 # XXX The Latex builder will t 162 # XXX The Latex builder will throw NoUri exceptions here,
166 # work around that by ignoring 163 # work around that by ignoring them.
167 # 164 #
168 try: 165 try:
169 xref = cdom.resolve_xref(a 166 xref = cdom.resolve_xref(app.env, docname, app.builder,
170 ' 167 'function', target, pxref,
171 l 168 lit_text)
172 except NoUri: 169 except NoUri:
173 xref = None 170 xref = None
174 171
175 if xref: 172 if xref:
176 return xref 173 return xref
177 note_failure(target) 174 note_failure(target)
178 175
179 return target_text 176 return target_text
180 177
181 def markup_c_ref(docname, app, match): 178 def markup_c_ref(docname, app, match):
182 class_str = {# Sphinx 2 only 179 class_str = {# Sphinx 2 only
183 RE_function: 'c-func', 180 RE_function: 'c-func',
184 RE_generic_type: 'c-type', 181 RE_generic_type: 'c-type',
185 # Sphinx 3+ only 182 # Sphinx 3+ only
186 RE_struct: 'c-struct', 183 RE_struct: 'c-struct',
187 RE_union: 'c-union', 184 RE_union: 'c-union',
188 RE_enum: 'c-enum', 185 RE_enum: 'c-enum',
189 RE_typedef: 'c-type', 186 RE_typedef: 'c-type',
190 } 187 }
191 reftype_str = {# Sphinx 2 only 188 reftype_str = {# Sphinx 2 only
192 RE_function: 'function', 189 RE_function: 'function',
193 RE_generic_type: 'type', 190 RE_generic_type: 'type',
194 # Sphinx 3+ only 191 # Sphinx 3+ only
195 RE_struct: 'struct', 192 RE_struct: 'struct',
196 RE_union: 'union', 193 RE_union: 'union',
197 RE_enum: 'enum', 194 RE_enum: 'enum',
198 RE_typedef: 'type', 195 RE_typedef: 'type',
199 } 196 }
200 197
201 cdom = app.env.domains['c'] 198 cdom = app.env.domains['c']
202 # 199 #
203 # Go through the dance of getting an xref 200 # Go through the dance of getting an xref out of the C domain
204 # 201 #
205 base_target = match.group(2) 202 base_target = match.group(2)
206 target_text = nodes.Text(match.group(0)) 203 target_text = nodes.Text(match.group(0))
207 xref = None 204 xref = None
208 possible_targets = [base_target] 205 possible_targets = [base_target]
209 # Check if this document has a namespace, 206 # Check if this document has a namespace, and if so, try
210 # cross-referencing inside it first. 207 # cross-referencing inside it first.
211 if c_namespace: 208 if c_namespace:
212 possible_targets.insert(0, c_namespace 209 possible_targets.insert(0, c_namespace + "." + base_target)
213 210
214 if base_target not in Skipnames: 211 if base_target not in Skipnames:
215 for target in possible_targets: 212 for target in possible_targets:
216 if not (match.re == RE_function an 213 if not (match.re == RE_function and target in Skipfuncs):
217 lit_text = nodes.literal(class 214 lit_text = nodes.literal(classes=['xref', 'c', class_str[match.re]])
218 lit_text += target_text 215 lit_text += target_text
219 pxref = addnodes.pending_xref( 216 pxref = addnodes.pending_xref('', refdomain = 'c',
220 217 reftype = reftype_str[match.re],
221 218 reftarget = target, modname = None,
222 219 classname = None)
223 # 220 #
224 # XXX The Latex builder will t 221 # XXX The Latex builder will throw NoUri exceptions here,
225 # work around that by ignoring 222 # work around that by ignoring them.
226 # 223 #
227 try: 224 try:
228 xref = cdom.resolve_xref(a 225 xref = cdom.resolve_xref(app.env, docname, app.builder,
229 r 226 reftype_str[match.re], target, pxref,
230 l 227 lit_text)
231 except NoUri: 228 except NoUri:
232 xref = None 229 xref = None
233 230
234 if xref: 231 if xref:
235 return xref 232 return xref
236 233
237 return target_text 234 return target_text
238 235
239 # 236 #
240 # Try to replace a documentation reference of 237 # Try to replace a documentation reference of the form Documentation/... with a
241 # cross reference to that page 238 # cross reference to that page
242 # 239 #
243 def markup_doc_ref(docname, app, match): 240 def markup_doc_ref(docname, app, match):
244 stddom = app.env.domains['std'] 241 stddom = app.env.domains['std']
245 # 242 #
246 # Go through the dance of getting an xref 243 # Go through the dance of getting an xref out of the std domain
247 # 244 #
248 absolute = match.group(1) 245 absolute = match.group(1)
249 target = match.group(2) 246 target = match.group(2)
250 if absolute: 247 if absolute:
251 target = "/" + target 248 target = "/" + target
252 xref = None 249 xref = None
253 pxref = addnodes.pending_xref('', refdomai 250 pxref = addnodes.pending_xref('', refdomain = 'std', reftype = 'doc',
254 reftarget = 251 reftarget = target, modname = None,
255 classname = 252 classname = None, refexplicit = False)
256 # 253 #
257 # XXX The Latex builder will throw NoUri e 254 # XXX The Latex builder will throw NoUri exceptions here,
258 # work around that by ignoring them. 255 # work around that by ignoring them.
259 # 256 #
260 try: 257 try:
261 xref = stddom.resolve_xref(app.env, do 258 xref = stddom.resolve_xref(app.env, docname, app.builder, 'doc',
262 target, pxr 259 target, pxref, None)
263 except NoUri: 260 except NoUri:
264 xref = None 261 xref = None
265 # 262 #
266 # Return the xref if we got it; otherwise 263 # Return the xref if we got it; otherwise just return the plain text.
267 # 264 #
268 if xref: 265 if xref:
269 return xref 266 return xref
270 else: 267 else:
271 return nodes.Text(match.group(0)) 268 return nodes.Text(match.group(0))
272 269
273 def get_c_namespace(app, docname): 270 def get_c_namespace(app, docname):
274 source = app.env.doc2path(docname) 271 source = app.env.doc2path(docname)
275 with open(source) as f: 272 with open(source) as f:
276 for l in f: 273 for l in f:
277 match = RE_namespace.search(l) 274 match = RE_namespace.search(l)
278 if match: 275 if match:
279 return match.group(1) 276 return match.group(1)
280 return '' 277 return ''
281 <<
282 def markup_git(docname, app, match): <<
283 # While we could probably assume that we a <<
284 # repository, we can't know for sure, so l <<
285 # turn them into git.kernel.org links with <<
286 # validity. (Maybe we can do something in <<
287 # these references if this is explicitly r <<
288 text = match.group(0) <<
289 rev = match.group('rev') <<
290 return nodes.reference('', nodes.Text(text <<
291 refuri=f'https://git.kernel.org/torval <<
292 278
293 def auto_markup(app, doctree, name): 279 def auto_markup(app, doctree, name):
294 global c_namespace 280 global c_namespace
295 c_namespace = get_c_namespace(app, name) 281 c_namespace = get_c_namespace(app, name)
296 def text_but_not_a_reference(node): 282 def text_but_not_a_reference(node):
297 # The nodes.literal test catches ``lit 283 # The nodes.literal test catches ``literal text``, its purpose is to
298 # avoid adding cross-references to fun 284 # avoid adding cross-references to functions that have been explicitly
299 # marked with cc:func:. 285 # marked with cc:func:.
300 if not isinstance(node, nodes.Text) or 286 if not isinstance(node, nodes.Text) or isinstance(node.parent, nodes.literal):
301 return False 287 return False
302 288
303 child_of_reference = False 289 child_of_reference = False
304 parent = node.parent 290 parent = node.parent
305 while parent: 291 while parent:
306 if isinstance(parent, nodes.Refere 292 if isinstance(parent, nodes.Referential):
307 child_of_reference = True 293 child_of_reference = True
308 break 294 break
309 parent = parent.parent 295 parent = parent.parent
310 return not child_of_reference 296 return not child_of_reference
311 297
312 # 298 #
313 # This loop could eventually be improved o 299 # This loop could eventually be improved on. Someday maybe we
314 # want a proper tree traversal with a lot 300 # want a proper tree traversal with a lot of awareness of which
315 # kinds of nodes to prune. But this works 301 # kinds of nodes to prune. But this works well for now.
316 # 302 #
317 for para in doctree.traverse(nodes.paragra 303 for para in doctree.traverse(nodes.paragraph):
318 for node in para.traverse(condition=te 304 for node in para.traverse(condition=text_but_not_a_reference):
319 node.parent.replace(node, markup_r 305 node.parent.replace(node, markup_refs(name, app, node))
320 306
321 def setup(app): 307 def setup(app):
322 app.connect('doctree-resolved', auto_marku 308 app.connect('doctree-resolved', auto_markup)
323 return { 309 return {
324 'parallel_read_safe': True, 310 'parallel_read_safe': True,
325 'parallel_write_safe': True, 311 'parallel_write_safe': True,
326 } 312 }
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.