Add symbol checking script to libc++ to help manage exported symbols.

Summary:
Add symbol checking scripts for extracting a list of symbols from shared libraries and for comparing symbol lists for differences.



Reviewers: mclow.lists, danalbert, EricWF

Reviewed By: EricWF

Subscribers: majnemer, emaste, cfe-commits

Differential Revision: http://reviews.llvm.org/D4946

git-svn-id: https://llvm.org/svn/llvm-project/libcxx/trunk@232855 91177308-0d34-0410-b5e6-96231b3b80d8

utils/sym_check/linux_blacklist.txt

@@ -0,0 +1,19 @@
+# all guard variables
+_ZGVNSt3__
+# all vtables
+_ZTV
+# all VTT
+_ZTT
+# all non-virtual thunks
+_ZTh
+# all virtual thunks
+_ZTv
+# typeinfo for std::__1::__types
+#    There are no std::__types
+_ZTINSt3__1[0-9][0-9]*__
+# typeinfo name for std::__1::__types
+_ZTSNSt3__1[0-9][0-9]*__
+# anything using __hidden_allocator
+.*__hidden_allocator
+# anything using __sso_allocator
+.*__sso_allocator

utils/sym_check/osx_blacklist.txt

@@ -0,0 +1,19 @@
+# all guard variables
+__ZGVNSt3__
+# all vtables
+__ZTV
+# all VTT
+__ZTT
+# all non-virtual thunks
+__ZTh
+# all virtual thunks
+__ZTv
+# typeinfo for std::__1::__types
+#    There are no std::__types
+__ZTINSt3__1[0-9][0-9]*__
+# typeinfo name for std::__1::__types
+__ZTSNSt3__1[0-9][0-9]*__
+# anything using __hidden_allocator
+.*__hidden_allocator
+# anything using __sso_allocator
+.*__sso_allocator

utils/sym_check/sym_check/__init__.py

@@ -0,0 +1,8 @@
+"""libcxx abi symbol checker"""
+
+__author__ = 'Eric Fiselier'
+__email__ = 'eric@efcs.ca'
+__versioninfo__ = (0, 1, 0)
+__version__ = ' '.join(str(v) for v in __versioninfo__) + 'dev'
+
+__all__ = ['diff', 'extract', 'util']

utils/sym_check/sym_check/diff.py

@@ -0,0 +1,93 @@
+# -*- Python -*- vim: set syntax=python tabstop=4 expandtab cc=80:
+"""
+diff - A set of functions for diff-ing two symbol lists.
+"""
+
+from sym_check import util
+
+
+def _symbol_difference(lhs, rhs):
+    lhs_names = set((n['name'] for n in lhs))
+    rhs_names = set((n['name'] for n in rhs))
+    diff_names = lhs_names - rhs_names
+    return [n for n in lhs if n['name'] in diff_names]
+
+
+def _find_by_key(sym_list, k):
+    for sym in sym_list:
+        if sym['name'] == k:
+            return sym
+    return None
+
+
+def added_symbols(old, new):
+    return _symbol_difference(new, old)
+
+
+def removed_symbols(old, new):
+    return _symbol_difference(old, new)
+
+
+def changed_symbols(old, new):
+    changed = []
+    for old_sym in old:
+        if old_sym in new:
+            continue
+        new_sym = _find_by_key(new, old_sym['name'])
+        if (new_sym is not None and not new_sym in old
+                and cmp(old_sym, new_sym) != 0):
+            changed += [(old_sym, new_sym)]
+    return changed
+
+
+def diff(old, new):
+    added = added_symbols(old, new)
+    removed = removed_symbols(old, new)
+    changed = changed_symbols(old, new)
+    return added, removed, changed
+
+
+def report_diff(added_syms, removed_syms, changed_syms, names_only=False,
+                demangle=True):
+    def maybe_demangle(name):
+        return util.demangle_symbol(name) if demangle else name
+
+    report = ''
+    for sym in added_syms:
+        report += 'Symbol added: %s\n' % maybe_demangle(sym['name'])
+        if not names_only:
+            report += '    %s\n\n' % sym
+    if added_syms and names_only:
+        report += '\n'
+    for sym in removed_syms:
+        report += 'SYMBOL REMOVED: %s\n' % maybe_demangle(sym['name'])
+        if not names_only:
+            report += '    %s\n\n' % sym
+    if removed_syms and names_only:
+        report += '\n'
+    if not names_only:
+        for sym_pair in changed_syms:
+            old_sym, new_sym = sym_pair
+            old_str = '\n    OLD SYMBOL: %s' % old_sym
+            new_str = '\n    NEW SYMBOL: %s' % new_sym
+            report += ('SYMBOL CHANGED: %s%s%s\n\n' %
+                       (maybe_demangle(old_sym['name']),
+                        old_str, new_str))
+
+    added = bool(len(added_syms) != 0)
+    abi_break = bool(len(removed_syms))
+    if not names_only:
+        abi_break = abi_break or len(changed_syms)
+    if added or abi_break:
+        report += 'Summary\n'
+        report += '    Added:   %d\n' % len(added_syms)
+        report += '    Removed: %d\n' % len(removed_syms)
+        if not names_only:
+            report += '    Changed: %d\n' % len(changed_syms)
+        if not abi_break:
+            report += 'Symbols added.'
+        else:
+            report += 'ABI BREAKAGE: SYMBOLS ADDED OR REMOVED!'
+    else:
+        report += 'Symbols match.'
+    return report, int(abi_break)

utils/sym_check/sym_check/extract.py

@@ -0,0 +1,101 @@
+# -*- Python -*- vim: set syntax=python tabstop=4 expandtab cc=80:
+"""
+extract - A set of function that extract symbol lists from shared libraries.
+"""
+import distutils.spawn
+import sys
+
+from sym_check import util
+
+
+class Extractor(object):
+    """
+    Extractor - Extract symbol lists from libraries using nm.
+    """
+
+    @staticmethod
+    def find_nm():
+        """
+        Search for the nm executable and return the path and type.
+        """
+        nm_exe = distutils.spawn.find_executable('nm')
+        if nm_exe is not None:
+            return nm_exe
+        # ERROR no NM found
+        print("ERROR: Could not find nm")
+        sys.exit(1)
+
+    def __init__(self):
+        """
+        Initialize the nm executable and flags that will be used to extract
+        symbols from shared libraries.
+        """
+        self.nm_exe = Extractor.find_nm()
+        self.flags = ['-P', '-g']
+
+    def extract(self, lib):
+        """
+        Extract symbols from a library and return the results as a dict of
+        parsed symbols.
+        """
+        cmd = [self.nm_exe] + self.flags + [lib]
+        out, _, exit_code = util.execute_command_verbose(cmd)
+        if exit_code != 0:
+            raise RuntimeError('Failed to run %s on %s' % (self.nm_exe, lib))
+        fmt_syms = (self._extract_sym(l)
+                    for l in out.splitlines() if l.strip())
+            # Cast symbol to string.
+        final_syms = (repr(s) for s in fmt_syms if self._want_sym(s))
+        # Make unique and sort strings.
+        tmp_list = list(sorted(set(final_syms)))
+        # Cast string back to symbol.
+        return util.read_syms_from_list(tmp_list)
+
+    def _extract_sym(self, sym_str):
+        bits = sym_str.split()
+        # Everything we want has at least two columns.
+        if len(bits) < 2:
+            return None
+        new_sym = {
+            'name': bits[0],
+            'type': bits[1]
+        }
+        new_sym = self._transform_sym_type(new_sym)
+        # NM types which we want to save the size for.
+        if new_sym['type'] == 'OBJECT' and len(bits) > 3:
+            new_sym['size'] = int(bits[3], 16)
+        return new_sym
+
+    @staticmethod
+    def _want_sym(sym):
+        """
+        Check that s is a valid symbol that we want to keep.
+        """
+        if sym is None or len(sym) < 2:
+            return False
+        bad_types = ['t', 'b', 'r', 'd', 'w']
+        return sym['type'] not in bad_types
+
+    @staticmethod
+    def _transform_sym_type(sym):
+        """
+        Map the nm single letter output for type to either FUNC or OBJECT.
+        If the type is not recognized it is left unchanged.
+        """
+        func_types = ['T', 'W']
+        obj_types = ['B', 'D', 'R', 'V', 'S']
+        if sym['type'] in func_types:
+            sym['type'] = 'FUNC'
+        elif sym['type'] in obj_types:
+            sym['type'] = 'OBJECT'
+        return sym
+
+
+def extract_symbols(lib_file):
+    """
+    Extract and return a list of symbols extracted from a dynamic library.
+    The symbols are extracted using NM. They are then filtered and formated.
+    Finally they symbols are made unique.
+    """
+    extractor = Extractor()
+    return extractor.extract(lib_file)

utils/sym_check/sym_check/match.py

@@ -0,0 +1,32 @@
+# -*- Python -*- vim: set syntax=python tabstop=4 expandtab cc=80:
+"""
+match - A set of functions for matching symbols in a list to a list of regexs
+"""
+
+import re
+
+
+def find_and_report_matching(symbol_list, regex_list):
+    report = ''
+    found_count = 0
+    for regex_str in regex_list:
+        report += 'Matching regex "%s":\n' % regex_str
+        matching_list = find_matching_symbols(symbol_list, regex_str)
+        if not matching_list:
+            report += '    No matches found\n\n'
+            continue
+        # else
+        found_count += len(matching_list)
+        for m in matching_list:
+            report += '    MATCHES: %s\n' % m['name']
+        report += '\n'
+    return found_count, report
+
+
+def find_matching_symbols(symbol_list, regex_str):
+    regex = re.compile(regex_str)
+    matching_list = []
+    for s in symbol_list:
+        if regex.match(s['name']):
+            matching_list += [s]
+    return matching_list

utils/sym_check/sym_check/util.py

@@ -0,0 +1,102 @@
+import ast
+import distutils.spawn
+import signal
+import subprocess
+import sys
+
+
+def execute_command(cmd, input_str=None):
+    """
+    Execute a command, capture and return its output.
+    """
+    kwargs = {
+        'stdin': subprocess.PIPE,
+        'stdout': subprocess.PIPE,
+        'stderr': subprocess.PIPE,
+    }
+    p = subprocess.Popen(cmd, **kwargs)
+    out, err = p.communicate(input=input_str)
+    exitCode = p.wait()
+    if exitCode == -signal.SIGINT:
+        raise KeyboardInterrupt
+    return out, err, exitCode
+
+
+def execute_command_verbose(cmd, input_str=None):
+    """
+    Execute a command and print its output on failure.
+    """
+    out, err, exitCode = execute_command(cmd, input_str=input_str)
+    if exitCode != 0:
+        report = "Command: %s\n" % ' '.join(["'%s'" % a for a in cmd])
+        report += "Exit Code: %d\n" % exitCode
+        if out:
+            report += "Standard Output:\n--\n%s--" % out
+        if err:
+            report += "Standard Error:\n--\n%s--" % err
+        report += "\n\nFailed!"
+        sys.stderr.write('%s\n' % report)
+    return out, err, exitCode
+
+
+def read_syms_from_list(slist):
+    """
+    Read a list of symbols from a list of strings.
+    Each string is one symbol.
+    """
+    return [ast.literal_eval(l) for l in slist]
+
+
+def read_syms_from_file(filename):
+    """
+    Read a list of symbols in from a file.
+    """
+    with open(filename, 'r') as f:
+        data = f.read()
+    return read_syms_from_list(data.splitlines())
+
+
+def read_blacklist(filename):
+    with open(filename, 'r') as f:
+        data = f.read()
+    lines = [l.strip() for l in data.splitlines() if l.strip()]
+    lines = [l for l in lines if not l.startswith('#')]
+    return lines
+
+
+def write_syms(sym_list, out=None, names_only=False):
+    """
+    Write a list of symbols to the file named by out.
+    """
+    out_str = ''
+    out_list = sym_list
+    if names_only:
+        out_list = [sym['name'] for sym in sym_list]
+        out_list.sort()
+    for sym in out_list:
+        out_str += '%s\n' % sym
+    if out is None:
+        sys.stdout.write(out_str)
+    else:
+        with open(out, 'w') as f:
+            f.write(out_str)
+
+
+_cppfilt_exe = distutils.spawn.find_executable('c++filt')
+
+
+def demangle_symbol(symbol):
+    if _cppfilt_exe is None:
+        return symbol
+    out, _, exit_code = execute_command_verbose(
+        [_cppfilt_exe], input_str=symbol)
+    if exit_code != 0:
+        return symbol
+    return out
+
+
+def extract_or_load(filename):
+    import sym_check.extract
+    if filename.endswith('.so') or filename.endswith('.dylib'):
+        return extract.extract_symbols(filename)
+    return read_syms_from_file(filename)

utils/sym_check/sym_diff.py

@@ -0,0 +1,47 @@
+#!/usr/bin/env python
+"""
+sym_diff - Compare two symbol lists and output the differences.
+"""
+from argparse import ArgumentParser
+import sys
+from sym_check import diff, util
+
+
+def main():
+    parser = ArgumentParser(
+        description='Extract a list of symbols from a shared library.')
+    parser.add_argument(
+        '--names-only', dest='names_only',
+        help='Only print symbol names',
+        action='store_true', default=False)
+    parser.add_argument(
+        '-o', '--output', dest='output',
+        help='The output file. stdout is used if not given',
+        type=str, action='store', default=None)
+    parser.add_argument(
+        '--demangle', dest='demangle', action='store_true', default=False)
+    parser.add_argument(
+        'old_syms', metavar='old-syms', type=str,
+        help='The file containing the old symbol list or a library')
+    parser.add_argument(
+        'new_syms', metavar='new-syms', type=str,
+        help='The file containing the new symbol list or a library')
+    args = parser.parse_args()
+
+    old_syms_list = util.extract_or_load(args.old_syms)
+    new_syms_list = util.extract_or_load(args.new_syms)
+
+    added, removed, changed = diff.diff(old_syms_list, new_syms_list)
+    report, is_break = diff.report_diff(added, removed, changed,
+                                        names_only=args.names_only,
+                                        demangle=args.demangle)
+    if args.output is None:
+        print(report)
+    else:
+        with open(args.output, 'w') as f:
+            f.write(report + '\n')
+    sys.exit(is_break)
+
+
+if __name__ == '__main__':
+    main()

utils/sym_check/sym_extract.py

@@ -0,0 +1,29 @@
+#!/usr/bin/env python
+"""
+sym_extract - Extract and output a list of symbols from a shared library.
+"""
+from argparse import ArgumentParser
+from sym_check import extract, util
+
+
+def main():
+    parser = ArgumentParser(
+        description='Extract a list of symbols from a shared library.')
+    parser.add_argument('library', metavar='shared-lib', type=str,
+                        help='The library to extract symbols from')
+    parser.add_argument('-o', '--output', dest='output',
+                        help='The output file. stdout is used if not given',
+                        type=str, action='store', default=None)
+    parser.add_argument('--names-only', dest='names_only',
+                        help='Output only the name of the symbol',
+                        action='store_true', default=False)
+    args = parser.parse_args()
+    if args.output is not None:
+        print('Extracting symbols from %s to %s.'
+              % (args.library, args.output))
+    syms = extract.extract_symbols(args.library)
+    util.write_syms(syms, out=args.output, names_only=args.names_only)
+
+
+if __name__ == '__main__':
+    main()

utils/sym_check/sym_match.py

@@ -0,0 +1,42 @@
+#!/usr/bin/env python
+"""
+sym_match - Match all symbols in a list against a list of regexes.
+"""
+from argparse import ArgumentParser
+import sys
+from sym_check import util, match, extract
+
+
+def main():
+    parser = ArgumentParser(
+        description='Extract a list of symbols from a shared library.')
+    parser.add_argument(
+        '--blacklist', dest='blacklist',
+        type=str, action='store', default=None)
+    parser.add_argument(
+        'symbol_list', metavar='symbol_list', type=str,
+        help='The file containing the old symbol list')
+    parser.add_argument(
+        'regexes', metavar='regexes', default=[], nargs='*',
+        help='The file containing the new symbol list or a library')
+    args = parser.parse_args()
+
+    if not args.regexes and args.blacklist is None:
+        sys.stderr.write('Either a regex or a blacklist must be specified.\n')
+        sys.exit(1)
+    if args.blacklist:
+        search_list = util.read_blacklist(args.blacklist)
+    else:
+        search_list = args.regexes
+
+    symbol_list = util.extract_or_load(args.symbol_list)
+
+    matching_count, report = match.find_and_report_matching(
+        symbol_list, search_list)
+    sys.stdout.write(report)
+    if matching_count != 0:
+        print('%d matching symbols found...' % matching_count)
+
+
+if __name__ == '__main__':
+    main()