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
|
import argparse
import sys, os
import base64
import hashlib
import tempfile
import shutil
import subprocess
from collections import namedtuple
from unittest.mock import patch
def call(cmd, cwd=None, env=None):
return subprocess.check_output(cmd, stdin=None, stderr=sys.stderr, cwd=cwd, env=env)
class LLVMResourceInserter:
def __init__(self, args, obj_output):
self.cxx = args.compiler
self.objcopy = args.objcopy
self.rescompiler = args.rescompiler
self.compressor = args.compressor
self.obj_out = obj_output
has_target = args.target or args.target != '__unknown_target__'
self.target_flags = [f"--target={args.target}"] if has_target else []
seed = os.path.basename(obj_output)
self.mangler = lambda key: 'R' + hashlib.sha256((seed + key).encode()).hexdigest()
self.tmpdir = tempfile.mkdtemp()
self.LARGE_RESOURCE_SIZE_THRESHOLD = args.large_resource_thr
self.SECTION_NAME = '__DATA,__res_holder' if 'apple' in args.target else '.rodata.__res_holder'
self.LARGE_SECTION_NAME = '__DATA,__res_holder' if 'apple' in args.target else '.lrodata.__res_holder'
# Actually there is no large sections on 32-bit arch
if args.arch_32_bits:
self.LARGE_SECTION_NAME = self.SECTION_NAME
def __enter__(self):
return self
def __exit__(self, *_):
shutil.rmtree(self.tmpdir)
def _get_section_name(self, size: int) -> str:
# If the inserted resource is larger (or equal) than X Mb, we will put it in the lrodata section.
LARGE_RESOURCE_SIZE_MB = self.LARGE_RESOURCE_SIZE_THRESHOLD
return self.LARGE_SECTION_NAME if (size >> 20) >= LARGE_RESOURCE_SIZE_MB else self.SECTION_NAME
PackedArgs = namedtuple("PackedArgs", ["infile_path", "comressed_file_path", "symbol_name"])
def _compress_files(self, infos: list[PackedArgs]) -> list[int]:
if len(infos) == 0:
return [0]
# Compress resources
cmd = [self.compressor, "--compress-only"]
for inserted_file, compressed_file, _ in infos:
cmd.extend([inserted_file, compressed_file])
call(cmd)
# Return sizes of compressed files
return [os.path.getsize(zipped_file) for (_, zipped_file, _) in infos]
@staticmethod
def flat_merge_obj(outs, output):
if len(outs) == 1:
shutil.move(outs[0], output)
return
with open(output, 'wb') as fout:
for fname in outs:
with open(fname, 'rb') as fin:
shutil.copyfileobj(fin, fout)
return
def _insert_files(self, infos: list[PackedArgs], f_sizes: list[int], output_obj: str) -> None:
TOTAL_SIZE = sum(f_sizes)
if len(infos) == 0:
return
# Merge files into one
compressed_files = [compressed_file for _, compressed_file, _ in infos]
merged_output = os.path.join(self.tmpdir, 'merged.zstd')
self.flat_merge_obj(compressed_files, merged_output)
# Update section content
SEC = self._get_section_name(TOTAL_SIZE)
cmd = [self.objcopy, f'--update-section={SEC}={merged_output}', output_obj]
call(cmd)
def _gen_prefix(self, infos: list[PackedArgs], f_sizes: list[int]) -> str:
TOTAL_SIZE = sum(f_sizes)
if len(infos) == 0 or TOTAL_SIZE == 0:
return ''
syms = []
for packed_args in infos:
syms += [sym for (_, _, sym) in packed_args]
cumulative_pos = [0] * (len(f_sizes) + 1)
for i, sz in enumerate(f_sizes):
cumulative_pos[i + 1] = cumulative_pos[i] + sz
cumulative_pos.pop()
SEC = self._get_section_name(TOTAL_SIZE)
SECTION_ATTR = f'__attribute__((section("{SEC}")))'
HOLDER_VARIABLE = 'resource_holder'
generated_return = f'static {SECTION_ATTR} const char {HOLDER_VARIABLE}[{TOTAL_SIZE}] = {{0}};\n'
template = '#define {sym} ({var} + {begin_offset})\n'
template += '#define {sym}_end ({var} + {end_offset})\n'
for sym, start_offset, size in zip(syms, cumulative_pos, f_sizes):
generated_return += template.format(
sym=sym, begin_offset=start_offset, end_offset=start_offset + size, var=HOLDER_VARIABLE
)
return generated_return
def insert_resources(self, kv_files, kv_strings):
kv_files = list(kv_files)
# Step 1: Compress files and save its sizes
file_sizes = []
infos = [[]]
estimated_cmd_len = 0
LIMIT = 6000
for idx, (path, key) in enumerate(kv_files):
packed_args = (path, os.path.join(self.tmpdir, f'{idx}.zstd'), self.mangler(key))
infos[-1].append(packed_args)
estimated_cmd_len += len(path)
if estimated_cmd_len > LIMIT:
infos.append([])
estimated_cmd_len = 0
for packed_args in infos:
file_sizes += self._compress_files(packed_args)
# Introduce custom tmp_filename generator for keeping object-file hash the same
def tmp_file_generator():
idx = 0
base = os.path.basename(self.obj_out)
while True:
yield f'{base}_{str(idx)}'
idx += 1
# Step 2: Generate resource registration cpp code & compile it
with patch.object(tempfile, "_get_candidate_names", tmp_file_generator), tempfile.NamedTemporaryFile(
suffix='.cc'
) as dummy_src:
cmd = [self.rescompiler, dummy_src.name, '--use-sections']
for path, key in kv_files + list(('-', k) for k in kv_strings):
if path != '-':
path = self.mangler(key)
cmd.extend([path, key])
call(cmd)
# Generate code for real char arrays, that will hold the resource
generated_source_text = ''
with open(dummy_src.name, 'rt') as f:
generated_source_text = f.read()
generated_source_text = generated_source_text.replace('extern "C" const char', '// extern "C" const char')
generated_source_text = self._gen_prefix(infos, file_sizes) + generated_source_text
with open(dummy_src.name, 'wt') as f:
f.write(generated_source_text)
# Compile
call([self.cxx, dummy_src.name, *self.target_flags, '-c', '-o', self.obj_out])
# Step 3: Put files into object
if sum(file_sizes) > 0:
self._insert_files(packed_args, file_sizes, self.obj_out)
return self.obj_out
def parse_args():
parser = argparse.ArgumentParser()
parser.add_argument('--compiler', required=True)
parser.add_argument('--objcopy', required=True)
parser.add_argument('--compressor', required=True)
parser.add_argument('--rescompiler', required=True)
parser.add_argument('--output_obj', required=True)
parser.add_argument('--inputs', nargs='+', required=False, default=[])
parser.add_argument('--keys', nargs='+', required=False, default=[])
parser.add_argument('--kvs', nargs='+', required=False, default=[])
parser.add_argument('--target', required=True)
parser.add_argument('--arch_32_bits', required=False, action='store_true', default=False)
help_msg = 'The threshold for large resources in megabytes (Mb). '
help_msg += 'If the compressed resource exceeds X Mb, it will be stored in the .lrodata '
help_msg += 'section instead of the .rodata section. This feature only works on 64-bit Linux systems. '
help_msg += 'This options can be changed via YMAKE_OBJCOPY_LARGE_RESOURCE_THR.'
parser.add_argument('--large_resource_thr', required=False, type=int, default=64, help=help_msg)
args = parser.parse_args()
# Decode hex to original string
args.keys = list(base64.b64decode(it).decode("utf-8") for it in args.keys)
return args, args.inputs, args.keys
def main():
args, inputs, keys = parse_args()
with LLVMResourceInserter(args, args.output_obj) as inserter:
inserter.insert_resources(zip(inputs, keys), args.kvs)
return
if __name__ == '__main__':
main()
|
