| 1 |
#!/usr/bin/env python3 |
| 2 |
|
| 3 |
""" |
| 4 |
strip_asm.py - Cleanup ASM output for the specified file |
| 5 |
""" |
| 6 |
|
| 7 |
import os |
| 8 |
import re |
| 9 |
import sys |
| 10 |
from argparse import ArgumentParser |
| 11 |
|
| 12 |
|
| 13 |
def find_used_labels(asm): |
| 14 |
found = set() |
| 15 |
label_re = re.compile(r"\s*j[a-z]+\s+\.L([a-zA-Z0-9][a-zA-Z0-9_]*)") |
| 16 |
for line in asm.splitlines(): |
| 17 |
m = label_re.match(line) |
| 18 |
if m: |
| 19 |
found.add(".L%s" % m.group(1)) |
| 20 |
return found |
| 21 |
|
| 22 |
|
| 23 |
def normalize_labels(asm): |
| 24 |
decls = set() |
| 25 |
label_decl = re.compile("^[.]{0,1}L([a-zA-Z0-9][a-zA-Z0-9_]*)(?=:)") |
| 26 |
for line in asm.splitlines(): |
| 27 |
m = label_decl.match(line) |
| 28 |
if m: |
| 29 |
decls.add(m.group(0)) |
| 30 |
if len(decls) == 0: |
| 31 |
return asm |
| 32 |
needs_dot = next(iter(decls))[0] != "." |
| 33 |
if not needs_dot: |
| 34 |
return asm |
| 35 |
for ld in decls: |
| 36 |
asm = re.sub(r"(^|\s+)" + ld + r"(?=:|\s)", "\\1." + ld, asm) |
| 37 |
return asm |
| 38 |
|
| 39 |
|
| 40 |
def transform_labels(asm): |
| 41 |
asm = normalize_labels(asm) |
| 42 |
used_decls = find_used_labels(asm) |
| 43 |
new_asm = "" |
| 44 |
label_decl = re.compile(r"^\.L([a-zA-Z0-9][a-zA-Z0-9_]*)(?=:)") |
| 45 |
for line in asm.splitlines(): |
| 46 |
m = label_decl.match(line) |
| 47 |
if not m or m.group(0) in used_decls: |
| 48 |
new_asm += line |
| 49 |
new_asm += "\n" |
| 50 |
return new_asm |
| 51 |
|
| 52 |
|
| 53 |
def is_identifier(tk): |
| 54 |
if len(tk) == 0: |
| 55 |
return False |
| 56 |
first = tk[0] |
| 57 |
if not first.isalpha() and first != "_": |
| 58 |
return False |
| 59 |
for i in range(1, len(tk)): |
| 60 |
c = tk[i] |
| 61 |
if not c.isalnum() and c != "_": |
| 62 |
return False |
| 63 |
return True |
| 64 |
|
| 65 |
|
| 66 |
def process_identifiers(line): |
| 67 |
""" |
| 68 |
process_identifiers - process all identifiers and modify them to have |
| 69 |
consistent names across all platforms; specifically across ELF and MachO. |
| 70 |
For example, MachO inserts an additional understore at the beginning of |
| 71 |
names. This function removes that. |
| 72 |
""" |
| 73 |
parts = re.split(r"([a-zA-Z0-9_]+)", line) |
| 74 |
new_line = "" |
| 75 |
for tk in parts: |
| 76 |
if is_identifier(tk): |
| 77 |
if tk.startswith("__Z"): |
| 78 |
tk = tk[1:] |
| 79 |
elif ( |
| 80 |
tk.startswith("_") |
| 81 |
and len(tk) > 1 |
| 82 |
and tk[1].isalpha() |
| 83 |
and tk[1] != "Z" |
| 84 |
): |
| 85 |
tk = tk[1:] |
| 86 |
new_line += tk |
| 87 |
return new_line |
| 88 |
|
| 89 |
|
| 90 |
def process_asm(asm): |
| 91 |
""" |
| 92 |
Strip the ASM of unwanted directives and lines |
| 93 |
""" |
| 94 |
new_contents = "" |
| 95 |
asm = transform_labels(asm) |
| 96 |
|
| 97 |
# TODO: Add more things we want to remove |
| 98 |
discard_regexes = [ |
| 99 |
re.compile(r"\s+\..*$"), # directive |
| 100 |
re.compile(r"\s*#(NO_APP|APP)$"), # inline ASM |
| 101 |
re.compile(r"\s*#.*$"), # comment line |
| 102 |
re.compile( |
| 103 |
r"\s*\.globa?l\s*([.a-zA-Z_][a-zA-Z0-9$_.]*)" |
| 104 |
), # global directive |
| 105 |
re.compile( |
| 106 |
r"\s*\.(string|asciz|ascii|[1248]?byte|short|word|long|quad|value|zero)" |
| 107 |
), |
| 108 |
] |
| 109 |
keep_regexes: list[re.Pattern] = [] |
| 110 |
fn_label_def = re.compile("^[a-zA-Z_][a-zA-Z0-9_.]*:") |
| 111 |
for line in asm.splitlines(): |
| 112 |
# Remove Mach-O attribute |
| 113 |
line = line.replace("@GOTPCREL", "") |
| 114 |
add_line = True |
| 115 |
for reg in discard_regexes: |
| 116 |
if reg.match(line) is not None: |
| 117 |
add_line = False |
| 118 |
break |
| 119 |
for reg in keep_regexes: |
| 120 |
if reg.match(line) is not None: |
| 121 |
add_line = True |
| 122 |
break |
| 123 |
if add_line: |
| 124 |
if fn_label_def.match(line) and len(new_contents) != 0: |
| 125 |
new_contents += "\n" |
| 126 |
line = process_identifiers(line) |
| 127 |
new_contents += line |
| 128 |
new_contents += "\n" |
| 129 |
return new_contents |
| 130 |
|
| 131 |
|
| 132 |
def main(): |
| 133 |
parser = ArgumentParser(description="generate a stripped assembly file") |
| 134 |
parser.add_argument( |
| 135 |
"input", |
| 136 |
metavar="input", |
| 137 |
type=str, |
| 138 |
nargs=1, |
| 139 |
help="An input assembly file", |
| 140 |
) |
| 141 |
parser.add_argument( |
| 142 |
"out", metavar="output", type=str, nargs=1, help="The output file" |
| 143 |
) |
| 144 |
args, unknown_args = parser.parse_known_args() |
| 145 |
input = args.input[0] |
| 146 |
output = args.out[0] |
| 147 |
if not os.path.isfile(input): |
| 148 |
print("ERROR: input file '%s' does not exist" % input) |
| 149 |
sys.exit(1) |
| 150 |
|
| 151 |
with open(input, "r") as f: |
| 152 |
contents = f.read() |
| 153 |
new_contents = process_asm(contents) |
| 154 |
with open(output, "w") as f: |
| 155 |
f.write(new_contents) |
| 156 |
|
| 157 |
|
| 158 |
if __name__ == "__main__": |
| 159 |
main() |
| 160 |
|
| 161 |
# vim: tabstop=4 expandtab shiftwidth=4 softtabstop=4 |
| 162 |
# kate: tab-width: 4; replace-tabs on; indent-width 4; tab-indents: off; |
| 163 |
# kate: indent-mode python; remove-trailing-spaces modified; |