expandlibs_exec.py

# This Source Code Form is subject to the terms of the Mozilla Public
# License, v. 2.0. If a copy of the MPL was not distributed with this
# file, You can obtain one at http://mozilla.org/MPL/2.0/.

'''expandlibs-exec.py applies expandlibs rules, and some more (see below) to
a given command line, and executes that command line with the expanded
arguments.

With the --extract argument (useful for e.g. $(AR)), it extracts object files
from static libraries (or use those listed in library descriptors directly).

With the --uselist argument (useful for e.g. $(CC)), it replaces all object
files with a list file. This can be used to avoid limitations in the length
of a command line. The kind of list file format used depends on the
EXPAND_LIBS_LIST_STYLE variable: 'list' for MSVC style lists (@file.list)
or 'linkerscript' for GNU ld linker scripts.
See https://bugzilla.mozilla.org/show_bug.cgi?id=584474#c59 for more details.

With the --symbol-order argument, followed by a file name, it will add the
relevant linker options to change the order in which the linker puts the
symbols appear in the resulting binary. Only works for ELF targets.
'''
from __future__ import with_statement
import sys
import os
from expandlibs import (
    ExpandArgs,
    relativize,
    isDynamicLib,
    isObject,
    ensureParentDir,
    ExpandLibsDeps,
)
import expandlibs_config as conf
from optparse import OptionParser
import subprocess
import tempfile
import shutil
import subprocess
import re
from mozbuild.makeutil import Makefile

# The are the insert points for a GNU ld linker script, assuming a more
# or less "standard" default linker script. This is not a dict because
# order is important.
SECTION_INSERT_BEFORE = [
  ('.text', '.fini'),
  ('.rodata', '.rodata1'),
  ('.data.rel.ro', '.dynamic'),
  ('.data', '.data1'),
]

class ExpandArgsMore(ExpandArgs):
    ''' Meant to be used as 'with ExpandArgsMore(args) as ...: '''
    def __enter__(self):
        self.tmp = []
        return self
        
    def __exit__(self, type, value, tb):
        '''Automatically remove temporary files'''
        for tmp in self.tmp:
            if os.path.isdir(tmp):
                shutil.rmtree(tmp, True)
            else:
                os.remove(tmp)

    def extract(self):
        self[0:] = self._extract(self)

    def _extract(self, args):
        '''When a static library name is found, either extract its contents
        in a temporary directory or use the information found in the
        corresponding lib descriptor.
        '''
        ar_extract = conf.AR_EXTRACT.split()
        newlist = []
        for arg in args:
            if os.path.splitext(arg)[1] == conf.LIB_SUFFIX:
                if os.path.exists(arg + conf.LIBS_DESC_SUFFIX):
                    newlist += self._extract(self._expand_desc(arg))
                    continue
                elif os.path.exists(arg) and (len(ar_extract) or conf.AR == 'lib'):
                    tmp = tempfile.mkdtemp(dir=os.curdir)
                    self.tmp.append(tmp)
                    if conf.AR == 'lib':
                        out = subprocess.check_output([conf.AR, '-NOLOGO', '-LIST', arg])
                        files = out.splitlines()
                        # If lib -list returns a list full of dlls, it's an
                        # import lib.
                        if all(isDynamicLib(f) for f in files):
                            newlist += [arg]
                            continue
                        for f in files:
                            subprocess.call([conf.AR, '-NOLOGO', '-EXTRACT:%s' % f, os.path.abspath(arg)], cwd=tmp)
                    else:
                        subprocess.call(ar_extract + [os.path.abspath(arg)], cwd=tmp)
                    objs = []
                    for root, dirs, files in os.walk(tmp):
                        objs += [relativize(os.path.join(root, f)) for f in files if isObject(f)]
                    newlist += sorted(objs)
                    continue
            newlist += [arg]
        return newlist

    def makelist(self):
        '''Replaces object file names with a temporary list file, using a
        list format depending on the EXPAND_LIBS_LIST_STYLE variable
        '''
        objs = [o for o in self if isObject(o)]
        if not len(objs): return
        fd, tmp = tempfile.mkstemp(suffix=".list",dir=os.curdir)
        if conf.EXPAND_LIBS_LIST_STYLE == "linkerscript":
            content = ['INPUT("%s")\n' % obj for obj in objs]
            ref = tmp
        elif conf.EXPAND_LIBS_LIST_STYLE == "filelist":
            content = ["%s\n" % obj for obj in objs]
            ref = "-Wl,-filelist," + tmp
        elif conf.EXPAND_LIBS_LIST_STYLE == "list":
            content = ["%s\n" % obj for obj in objs]
            ref = "@" + tmp
        else:
            os.close(fd)
            os.remove(tmp)
            return
        self.tmp.append(tmp)
        f = os.fdopen(fd, "w")
        f.writelines(content)
        f.close()
        idx = self.index(objs[0])
        newlist = self[0:idx] + [ref] + [item for item in self[idx:] if item not in objs]
        self[0:] = newlist

    def _getFoldedSections(self):
        '''Returns a dict about folded sections.
        When section A and B are folded into section C, the dict contains:
        { 'A': 'C',
          'B': 'C',
          'C': ['A', 'B'] }'''
        if not conf.LD_PRINT_ICF_SECTIONS:
            return {}

        proc = subprocess.Popen(self + [conf.LD_PRINT_ICF_SECTIONS], stdout = subprocess.PIPE, stderr = subprocess.PIPE)
        (stdout, stderr) = proc.communicate()
        result = {}
        # gold's --print-icf-sections output looks like the following:
        # ld: ICF folding section '.section' in file 'file.o'into '.section' in file 'file.o'
        # In terms of words, chances are this will change in the future,
        # especially considering "into" is misplaced. Splitting on quotes
        # seems safer.
        for l in stderr.split('\n'):
            quoted = l.split("'")
            if len(quoted) > 5 and quoted[1] != quoted[5]:
                result[quoted[1]] = [quoted[5]]
                if quoted[5] in result:
                    result[quoted[5]].append(quoted[1])
                else:
                    result[quoted[5]] = [quoted[1]]
        return result

    def _getOrderedSections(self, ordered_symbols):
        '''Given an ordered list of symbols, returns the corresponding list
        of sections following the order.'''
        if not conf.EXPAND_LIBS_ORDER_STYLE in ['linkerscript', 'section-ordering-file']:
            raise Exception('EXPAND_LIBS_ORDER_STYLE "%s" is not supported' % conf.EXPAND_LIBS_ORDER_STYLE)
        finder = SectionFinder([arg for arg in self if isObject(arg) or os.path.splitext(arg)[1] == conf.LIB_SUFFIX])
        folded = self._getFoldedSections()
        sections = set()
        ordered_sections = []
        for symbol in ordered_symbols:
            symbol_sections = finder.getSections(symbol)
            all_symbol_sections = []
            for section in symbol_sections:
                if section in folded:
                    if isinstance(folded[section], str):
                        section = folded[section]
                    all_symbol_sections.append(section)
                    all_symbol_sections.extend(folded[section])
                else:
                    all_symbol_sections.append(section)
            for section in all_symbol_sections:
                if not section in sections:
                    ordered_sections.append(section)
                    sections.add(section)
        return ordered_sections

    def orderSymbols(self, order):
        '''Given a file containing a list of symbols, adds the appropriate
        argument to make the linker put the symbols in that order.'''
        with open(order) as file:
            sections = self._getOrderedSections([l.strip() for l in file.readlines() if l.strip()])
        split_sections = {}
        linked_sections = [s[0] for s in SECTION_INSERT_BEFORE]
        for s in sections:
            for linked_section in linked_sections:
                if s.startswith(linked_section):
                    if linked_section in split_sections:
                        split_sections[linked_section].append(s)
                    else:
                        split_sections[linked_section] = [s]
                    break
        content = []
        # Order is important
        linked_sections = [s for s in linked_sections if s in split_sections]

        if conf.EXPAND_LIBS_ORDER_STYLE == 'section-ordering-file':
            option = '-Wl,--section-ordering-file,%s'
            content = sections
            for linked_section in linked_sections:
                content.extend(split_sections[linked_section])
                content.append('%s.*' % linked_section)
                content.append(linked_section)

        elif conf.EXPAND_LIBS_ORDER_STYLE == 'linkerscript':
            option = '-Wl,-T,%s'
            section_insert_before = dict(SECTION_INSERT_BEFORE)
            for linked_section in linked_sections:
                content.append('SECTIONS {')
                content.append('  %s : {' % linked_section)
                content.extend('    *(%s)' % s for s in split_sections[linked_section])
                content.append('  }')
                content.append('}')
                content.append('INSERT BEFORE %s' % section_insert_before[linked_section])
        else:
            raise Exception('EXPAND_LIBS_ORDER_STYLE "%s" is not supported' % conf.EXPAND_LIBS_ORDER_STYLE)

        fd, tmp = tempfile.mkstemp(dir=os.curdir)
        f = os.fdopen(fd, "w")
        f.write('\n'.join(content)+'\n')
        f.close()
        self.tmp.append(tmp)
        self.append(option % tmp)

class SectionFinder(object):
    '''Instances of this class allow to map symbol names to sections in
    object files.'''

    def __init__(self, objs):
        '''Creates an instance, given a list of object files.'''
        if not conf.EXPAND_LIBS_ORDER_STYLE in ['linkerscript', 'section-ordering-file']:
            raise Exception('EXPAND_LIBS_ORDER_STYLE "%s" is not supported' % conf.EXPAND_LIBS_ORDER_STYLE)
        self.mapping = {}
        for obj in objs:
            if not isObject(obj) and os.path.splitext(obj)[1] != conf.LIB_SUFFIX:
                raise Exception('%s is not an object nor a static library' % obj)
            for symbol, section in SectionFinder._getSymbols(obj):
                sym = SectionFinder._normalize(symbol)
                if sym in self.mapping:
                    if not section in self.mapping[sym]:
                        self.mapping[sym].append(section)
                else:
                    self.mapping[sym] = [section]

    def getSections(self, symbol):
        '''Given a symbol, returns a list of sections containing it or the
        corresponding thunks. When the given symbol is a thunk, returns the
        list of sections containing its corresponding normal symbol and the
        other thunks for that symbol.'''
        sym = SectionFinder._normalize(symbol)
        if sym in self.mapping:
            return self.mapping[sym]
        return []

    @staticmethod
    def _normalize(symbol):
        '''For normal symbols, return the given symbol. For thunks, return
        the corresponding normal symbol.'''
        if re.match('^_ZThn[0-9]+_', symbol):
            return re.sub('^_ZThn[0-9]+_', '_Z', symbol)
        return symbol

    @staticmethod
    def _getSymbols(obj):
        '''Returns a list of (symbol, section) contained in the given object
        file.'''
        proc = subprocess.Popen(['objdump', '-t', obj], stdout = subprocess.PIPE, stderr = subprocess.PIPE)
        (stdout, stderr) = proc.communicate()
        syms = []
        for line in stdout.splitlines():
            # Each line has the following format:
            # <addr> [lgu!][w ][C ][W ][Ii ][dD ][FfO ] <section>\t<length> <symbol>
            tmp = line.split(' ',1)
            # This gives us ["<addr>", "[lgu!][w ][C ][W ][Ii ][dD ][FfO ] <section>\t<length> <symbol>"]
            # We only need to consider cases where "<section>\t<length> <symbol>" is present,
            # and where the [FfO] flag is either F (function) or O (object).
            if len(tmp) > 1 and len(tmp[1]) > 6 and tmp[1][6] in ['O', 'F']:
                tmp = tmp[1][8:].split()
                # That gives us ["<section>","<length>", "<symbol>"]
                syms.append((tmp[-1], tmp[0]))
        return syms

def print_command(out, args):
    print >>out, "Executing: " + " ".join(args)
    for tmp in [f for f in args.tmp if os.path.isfile(f)]:
        print >>out, tmp + ":"
        with open(tmp) as file:
            print >>out, "".join(["    " + l for l in file.readlines()])
    out.flush()

def main():
    parser = OptionParser()
    parser.add_option("--depend", dest="depend", metavar="FILE",
        help="generate dependencies for the given execution and store it in the given file")
    parser.add_option("--target", dest="target", metavar="FILE",
        help="designate the target for dependencies")
    parser.add_option("--extract", action="store_true", dest="extract",
        help="when a library has no descriptor file, extract it first, when possible")
    parser.add_option("--uselist", action="store_true", dest="uselist",
        help="use a list file for objects when executing a command")
    parser.add_option("--verbose", action="store_true", dest="verbose",
        help="display executed command and temporary files content")
    parser.add_option("--symbol-order", dest="symbol_order", metavar="FILE",
        help="use the given list of symbols to order symbols in the resulting binary when using with a linker")

    (options, args) = parser.parse_args()

    if not options.target:
        options.depend = False
    if options.depend:
        deps = ExpandLibsDeps(args)
        # Filter out common command wrappers
        while os.path.basename(deps[0]) in ['ccache', 'distcc']:
            deps.pop(0)
        # Remove command
        deps.pop(0)
    with ExpandArgsMore(args) as args:
        if options.extract:
            args.extract()
        if options.symbol_order:
            args.orderSymbols(options.symbol_order)
        if options.uselist:
            args.makelist()

        if options.verbose:
            print_command(sys.stderr, args)
        try:
            proc = subprocess.Popen(args, stdout = subprocess.PIPE, stderr = subprocess.STDOUT)
        except Exception, e:
            print >>sys.stderr, 'error: Launching', args, ':', e
            raise e
        (stdout, stderr) = proc.communicate()
        if proc.returncode and not options.verbose:
            print_command(sys.stderr, args)
        sys.stderr.write(stdout)
        sys.stderr.flush()
        if proc.returncode:
            exit(proc.returncode)
    if not options.depend:
        return
    ensureParentDir(options.depend)
    mk = Makefile()
    deps = [dep for dep in deps if os.path.isfile(dep) and dep != options.target]
    no_dynamic_lib = [dep for dep in deps if not isDynamicLib(dep)]
    mk.create_rule([options.target]).add_dependencies(no_dynamic_lib)
    if len(deps) != len(no_dynamic_lib):
        mk.create_rule(['%s_order_only' % options.target]).add_dependencies(dep for dep in deps if isDynamicLib(dep))

    with open(options.depend, 'w') as depfile:
        mk.dump(depfile, removal_guard=True)

if __name__ == '__main__':
    main()