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
|
#!/usr/bin/env python
# encoding: utf-8
# Thomas Nagy, 2011 (ita)
"""
A make-like way of executing the build, following the relationships between inputs/outputs
This algorithm will lead to slower builds, will not be as flexible as "waf build", but
it might be useful for building data files (?)
It is likely to break in the following cases:
- files are created dynamically (no inputs or outputs)
- headers
- building two files from different groups
"""
import re
from waflib import Options, Task
from waflib.Build import BuildContext
class MakeContext(BuildContext):
'''executes tasks in a step-by-step manner, following dependencies between inputs/outputs'''
cmd = 'make'
fun = 'build'
def __init__(self, **kw):
super(MakeContext, self).__init__(**kw)
self.files = Options.options.files
def get_build_iterator(self):
if not self.files:
while 1:
yield super(MakeContext, self).get_build_iterator()
for g in self.groups:
for tg in g:
try:
f = tg.post
except AttributeError:
pass
else:
f()
provides = {}
uses = {}
all_tasks = []
tasks = []
for pat in self.files.split(','):
matcher = self.get_matcher(pat)
for tg in g:
if isinstance(tg, Task.Task):
lst = [tg]
else:
lst = tg.tasks
for tsk in lst:
all_tasks.append(tsk)
do_exec = False
for node in tsk.inputs:
try:
uses[node].append(tsk)
except:
uses[node] = [tsk]
if matcher(node, output=False):
do_exec = True
break
for node in tsk.outputs:
try:
provides[node].append(tsk)
except:
provides[node] = [tsk]
if matcher(node, output=True):
do_exec = True
break
if do_exec:
tasks.append(tsk)
# so we have the tasks that we need to process, the list of all tasks,
# the map of the tasks providing nodes, and the map of tasks using nodes
if not tasks:
# if there are no tasks matching, return everything in the current group
result = all_tasks
else:
# this is like a big filter...
result = set()
seen = set()
cur = set(tasks)
while cur:
result |= cur
tosee = set()
for tsk in cur:
for node in tsk.inputs:
if node in seen:
continue
seen.add(node)
tosee |= set(provides.get(node, []))
cur = tosee
result = list(result)
Task.set_file_constraints(result)
Task.set_precedence_constraints(result)
yield result
while 1:
yield []
def get_matcher(self, pat):
# this returns a function
inn = True
out = True
if pat.startswith('in:'):
out = False
pat = pat.replace('in:', '')
elif pat.startswith('out:'):
inn = False
pat = pat.replace('out:', '')
anode = self.root.find_node(pat)
pattern = None
if not anode:
if not pat.startswith('^'):
pat = '^.+?%s' % pat
if not pat.endswith('$'):
pat = '%s$' % pat
pattern = re.compile(pat)
def match(node, output):
if output and not out:
return False
if not output and not inn:
return False
if anode:
return anode == node
else:
return pattern.match(node.abspath())
return match
|