#!/usr/bin/env python # Copyright 2017 The Chromium Authors. All rights reserved. # Use of this source code is governed by a BSD-style license that can be # found in the LICENSE file. """Find header files missing in GN. This script gets all the header files from ninja_deps, which is from the true dependency generated by the compiler, and report if they don't exist in GN. """ import argparse import json import os import re import shutil import subprocess import sys import tempfile from multiprocessing import Process, Queue SRC_DIR = os.path.abspath( os.path.join(os.path.abspath(os.path.dirname(__file__)), os.path.pardir)) DEPOT_TOOLS_DIR = os.path.join(SRC_DIR, 'third_party', 'depot_tools') def GetHeadersFromNinja(out_dir, skip_obj, q): """Return all the header files from ninja_deps""" def NinjaSource(): cmd = [os.path.join(DEPOT_TOOLS_DIR, 'ninja'), '-C', out_dir, '-t', 'deps'] # A negative bufsize means to use the system default, which usually # means fully buffered. popen = subprocess.Popen(cmd, stdout=subprocess.PIPE, bufsize=-1) for line in iter(popen.stdout.readline, ''): yield line.rstrip() popen.stdout.close() return_code = popen.wait() if return_code: raise subprocess.CalledProcessError(return_code, cmd) ans, err = set(), None try: ans = ParseNinjaDepsOutput(NinjaSource(), out_dir, skip_obj) except Exception as e: err = str(e) q.put((ans, err)) def ParseNinjaDepsOutput(ninja_out, out_dir, skip_obj): """Parse ninja output and get the header files""" all_headers = {} # Ninja always uses "/", even on Windows. prefix = '../../' is_valid = False obj_file = '' for line in ninja_out: if line.startswith(' '): if not is_valid: continue if line.endswith('.h') or line.endswith('.hh'): f = line.strip() if f.startswith(prefix): f = f[6:] # Remove the '../../' prefix # build/ only contains build-specific files like build_config.h # and buildflag.h, and system header files, so they should be # skipped. if f.startswith(out_dir) or f.startswith('out'): continue if not f.startswith('build'): all_headers.setdefault(f, []) if not skip_obj: all_headers[f].append(obj_file) else: is_valid = line.endswith('(VALID)') obj_file = line.split(':')[0] return all_headers def GetHeadersFromGN(out_dir, q): """Return all the header files from GN""" tmp = None ans, err = set(), None try: # Argument |dir| is needed to make sure it's on the same drive on Windows. # dir='' means dir='.', but doesn't introduce an unneeded prefix. tmp = tempfile.mkdtemp(dir='') shutil.copy2(os.path.join(out_dir, 'args.gn'), os.path.join(tmp, 'args.gn')) # Do "gn gen" in a temp dir to prevent dirtying |out_dir|. gn_exe = 'gn.bat' if sys.platform == 'win32' else 'gn' subprocess.check_call([ os.path.join(DEPOT_TOOLS_DIR, gn_exe), 'gen', tmp, '--ide=json', '-q']) gn_json = json.load(open(os.path.join(tmp, 'project.json'))) ans = ParseGNProjectJSON(gn_json, out_dir, tmp) except Exception as e: err = str(e) finally: if tmp: shutil.rmtree(tmp) q.put((ans, err)) def ParseGNProjectJSON(gn, out_dir, tmp_out): """Parse GN output and get the header files""" all_headers = set() for _target, properties in gn['targets'].iteritems(): sources = properties.get('sources', []) public = properties.get('public', []) # Exclude '"public": "*"'. if type(public) is list: sources += public for f in sources: if f.endswith('.h') or f.endswith('.hh'): if f.startswith('//'): f = f[2:] # Strip the '//' prefix. if f.startswith(tmp_out): f = out_dir + f[len(tmp_out):] all_headers.add(f) return all_headers def GetDepsPrefixes(q): """Return all the folders controlled by DEPS file""" prefixes, err = set(), None try: gclient_exe = 'gclient.bat' if sys.platform == 'win32' else 'gclient' gclient_out = subprocess.check_output([ os.path.join(DEPOT_TOOLS_DIR, gclient_exe), 'recurse', '--no-progress', '-j1', 'python', '-c', 'import os;print os.environ["GCLIENT_DEP_PATH"]'], universal_newlines=True) for i in gclient_out.split('\n'): if i.startswith('src/'): i = i[4:] prefixes.add(i) except Exception as e: err = str(e) q.put((prefixes, err)) def IsBuildClean(out_dir): cmd = [os.path.join(DEPOT_TOOLS_DIR, 'ninja'), '-C', out_dir, '-n'] try: out = subprocess.check_output(cmd) return 'no work to do.' in out except Exception as e: print e return False def ParseWhiteList(whitelist): out = set() for line in whitelist.split('\n'): line = re.sub(r'#.*', '', line).strip() if line: out.add(line) return out def FilterOutDepsedRepo(files, deps): return {f for f in files if not any(f.startswith(d) for d in deps)} def GetNonExistingFiles(lst): out = set() for f in lst: if not os.path.isfile(f): out.add(f) return out def main(): def DumpJson(data): if args.json: with open(args.json, 'w') as f: json.dump(data, f) def PrintError(msg): DumpJson([]) parser.error(msg) parser = argparse.ArgumentParser(description=''' NOTE: Use ninja to build all targets in OUT_DIR before running this script.''') parser.add_argument('--out-dir', metavar='OUT_DIR', default='out/Release', help='output directory of the build') parser.add_argument('--json', help='JSON output filename for missing headers') parser.add_argument('--whitelist', help='file containing whitelist') parser.add_argument('--skip-dirty-check', action='store_true', help='skip checking whether the build is dirty') parser.add_argument('--verbose', action='store_true', help='print more diagnostic info') args, _extras = parser.parse_known_args() if not os.path.isdir(args.out_dir): parser.error('OUT_DIR "%s" does not exist.' % args.out_dir) if not args.skip_dirty_check and not IsBuildClean(args.out_dir): dirty_msg = 'OUT_DIR looks dirty. You need to build all there.' if args.json: # Assume running on the bots. Silently skip this step. # This is possible because "analyze" step can be wrong due to # underspecified header files. See crbug.com/725877 print dirty_msg DumpJson([]) return 0 else: # Assume running interactively. parser.error(dirty_msg) d_q = Queue() d_p = Process(target=GetHeadersFromNinja, args=(args.out_dir, True, d_q,)) d_p.start() gn_q = Queue() gn_p = Process(target=GetHeadersFromGN, args=(args.out_dir, gn_q,)) gn_p.start() deps_q = Queue() deps_p = Process(target=GetDepsPrefixes, args=(deps_q,)) deps_p.start() d, d_err = d_q.get() gn, gn_err = gn_q.get() missing = set(d.keys()) - gn nonexisting = GetNonExistingFiles(gn) deps, deps_err = deps_q.get() missing = FilterOutDepsedRepo(missing, deps) nonexisting = FilterOutDepsedRepo(nonexisting, deps) d_p.join() gn_p.join() deps_p.join() if d_err: PrintError(d_err) if gn_err: PrintError(gn_err) if deps_err: PrintError(deps_err) if len(GetNonExistingFiles(d)) > 0: print 'Non-existing files in ninja deps:', GetNonExistingFiles(d) PrintError('Found non-existing files in ninja deps. You should ' + 'build all in OUT_DIR.') if len(d) == 0: PrintError('OUT_DIR looks empty. You should build all there.') if any((('/gen/' in i) for i in nonexisting)): PrintError('OUT_DIR looks wrong. You should build all there.') if args.whitelist: whitelist = ParseWhiteList(open(args.whitelist).read()) missing -= whitelist nonexisting -= whitelist missing = sorted(missing) nonexisting = sorted(nonexisting) DumpJson(sorted(missing + nonexisting)) if len(missing) == 0 and len(nonexisting) == 0: return 0 if len(missing) > 0: print '\nThe following files should be included in gn files:' for i in missing: print i if len(nonexisting) > 0: print '\nThe following non-existing files should be removed from gn files:' for i in nonexisting: print i if args.verbose: # Only get detailed obj dependency here since it is slower. GetHeadersFromNinja(args.out_dir, False, d_q) d, d_err = d_q.get() print '\nDetailed dependency info:' for f in missing: print f for cc in d[f]: print ' ', cc print '\nMissing headers sorted by number of affected object files:' count = {k: len(v) for (k, v) in d.iteritems()} for f in sorted(count, key=count.get, reverse=True): if f in missing: print count[f], f return 1 if __name__ == '__main__': sys.exit(main())