]> git.decadent.org.uk Git - dak.git/commitdiff
Rewrite most of dak/contents.py.
authorTorsten Werner <twerner@debian.org>
Sat, 5 Mar 2011 17:26:52 +0000 (18:26 +0100)
committerTorsten Werner <twerner@debian.org>
Sat, 5 Mar 2011 17:26:52 +0000 (18:26 +0100)
Signed-off-by: Torsten Werner <twerner@debian.org>
dak/contents.py

index 99e9575d874d39ac0d4745b217ee529b2373037e..24146547e68828204e99b44b963acf81ea03ddf7 100755 (executable)
@@ -5,6 +5,7 @@ Create all the contents files
 @contact: Debian FTPMaster <ftpmaster@debian.org>
 @copyright: 2008, 2009 Michael Casadevall <mcasadevall@debian.org>
 @copyright: 2009 Mike O'Connor <stew@debian.org>
+@copyright: 2011 Torsten Werner <twerner@debian.org>
 @license: GNU General Public License version 2 or later
 """
 
@@ -35,667 +36,96 @@ Create all the contents files
 ################################################################################
 
 import sys
-import os
-import logging
-import gzip
-import threading
-import traceback
-import Queue
 import apt_pkg
-import datetime
-import traceback
-from daklib import utils
-from daklib.binary import Binary
+
 from daklib.config import Config
 from daklib.dbconn import *
 from daklib.contents import ContentsScanner
 from daklib import daklog
-
-Logger = None
+from daklib import utils
 
 ################################################################################
 
 def usage (exit_code=0):
-    print """Usage: dak contents [options] command [arguments]
+    print """Usage: dak contents [options] subcommand
 
-COMMANDS
+SUBCOMMANDS
     generate
         generate Contents-$arch.gz files
 
-    bootstrap_bin
+    scan
         scan the debs in the existing pool and load contents into the bin_contents table
 
-    bootstrap
-        copy data from the bin_contents table into the deb_contents / udeb_contents tables
-
-    cruft
-        remove files/paths which are no longer referenced by a binary
-
 OPTIONS
      -h, --help
         show this help and exit
 
-     -v, --verbose
-        show verbose information messages
-
-     -q, --quiet
-        supress all output but errors
-
+OPTIONS for generate
      -s, --suite={stable,testing,unstable,...}
-        only operate on a single suite
+        only operate on specified suite names
+
+     -f, --force
+        write Contents files for suites marked as untouchable, too
 
+OPTIONS for scan
      -l, --limit=NUMBER
-        optional package limit for bootstrap_bin
+        maximum number of packages to scan
 """
     sys.exit(exit_code)
 
 ################################################################################
 
-# where in dak.conf all of our configuration will be stowed
-
-options_prefix = "Contents"
-options_prefix = "%s::Options" % options_prefix
-
-log = logging.getLogger()
+def write_all(cnf, suite_names = [], force = None):
+    Logger = daklog.Logger(cnf.Cnf, 'contents generate')
+    result = ContentsScanner.scan_all(limit)
+    Logger.close()
 
 ################################################################################
 
-class EndOfContents(object):
-    """
-    A sentry object for the end of the filename stream
-    """
-    pass
-
-class OneAtATime(object):
-    """
-    a one space queue which sits between multiple possible producers
-    and multiple possible consumers
-    """
-    def __init__(self):
-        self.next_in_line = None
-        self.read_lock = threading.Condition()
-        self.write_lock = threading.Condition()
-        self.die = False
-
-    def enqueue(self, next):
-        self.write_lock.acquire()
-        while self.next_in_line:
-            if self.die:
-                return
-            self.write_lock.wait()
-
-        assert( not self.next_in_line )
-        self.next_in_line = next
-        self.write_lock.release()
-        self.read_lock.acquire()
-        self.read_lock.notify()
-        self.read_lock.release()
-
-    def dequeue(self):
-        self.read_lock.acquire()
-        while not self.next_in_line:
-            if self.die:
-                return
-            self.read_lock.wait()
-
-        result = self.next_in_line
-
-        self.next_in_line = None
-        self.read_lock.release()
-        self.write_lock.acquire()
-        self.write_lock.notify()
-        self.write_lock.release()
-
-        return result
-
-
-class ContentsWorkThread(threading.Thread):
-    """
-    """
-    def __init__(self, upstream, downstream):
-        threading.Thread.__init__(self)
-        self.upstream = upstream
-        self.downstream = downstream
-
-    def run(self):
-        while True:
-            try:
-                contents_file = self.upstream.dequeue()
-                if isinstance(contents_file,EndOfContents):
-                    if self.downstream:
-                        self.downstream.enqueue(contents_file)
-                    break
-
-                s = datetime.datetime.now()
-                print("%s start: %s" % (self,contents_file) )
-                self._run(contents_file)
-                print("%s finished: %s in %d seconds" % (self, contents_file, (datetime.datetime.now()-s).seconds ))
-                if self.downstream:
-                    self.downstream.enqueue(contents_file)
-            except:
-                traceback.print_exc()
-
-class QueryThread(ContentsWorkThread):
-    def __init__(self, upstream, downstream):
-        ContentsWorkThread.__init__(self, upstream, downstream)
-
-    def __str__(self):
-        return "QueryThread"
-    __repr__ = __str__
-
-    def _run(self, contents_file):
-        contents_file.query()
-
-class IngestThread(ContentsWorkThread):
-    def __init__(self, upstream, downstream):
-        ContentsWorkThread.__init__(self, upstream, downstream)
-
-    def __str__(self):
-        return "IngestThread"
-    __repr__ = __str__
-
-    def _run(self, contents_file):
-        contents_file.ingest()
-
-class SortThread(ContentsWorkThread):
-    def __init__(self, upstream, downstream):
-        ContentsWorkThread.__init__(self, upstream, downstream)
-
-    def __str__(self):
-        return "SortThread"
-    __repr__ = __str__
-
-    def _run(self, contents_file):
-        contents_file.sorted_keys = sorted(contents_file.filenames.keys())
-
-class OutputThread(ContentsWorkThread):
-    def __init__(self, upstream, downstream):
-        ContentsWorkThread.__init__(self, upstream, downstream)
-
-    def __str__(self):
-        return "OutputThread"
-    __repr__ = __str__
-
-    def _run(self, contents_file):
-        contents_file.open_file()
-        for fname in contents_file.sorted_keys:
-            contents_file.filehandle.write("%s\t%s\n" % (fname,contents_file.filenames[fname]))
-        contents_file.sorted_keys = None
-        contents_file.filenames.clear()
-
-class GzipThread(ContentsWorkThread):
-    def __init__(self, upstream, downstream):
-        ContentsWorkThread.__init__(self, upstream, downstream)
-
-    def __str__(self):
-        return "GzipThread"
-    __repr__ = __str__
-
-    def _run(self, contents_file):
-        os.system("gzip -f %s" % contents_file.filename)
-
-class ContentFile(object):
-    def __init__(self,
-                 filename,
-                 suite_str,
-                 suite_id):
-
-        self.filename = filename
-        self.filenames = {}
-        self.sorted_keys = None
-        self.suite_str = suite_str
-        self.suite_id = suite_id
-        self.session = None
-        self.filehandle = None
-        self.results = None
-
-    def __str__(self):
-        return self.filename
-    __repr__ = __str__
-
-
-    def cleanup(self):
-        self.filenames = None
-        self.sortedkeys = None
-        self.filehandle.close()
-        self.session.close()
-
-    def ingest(self):
-        while True:
-            r = self.results.fetchone()
-            if not r:
-                break
-            filename, package = r
-            self.filenames[filename]=package
-
-        self.session.close()
-
-    def open_file(self):
-        """
-        opens a gzip stream to the contents file
-        """
-        filepath = Config()["Contents::Root"] + self.filename
-        filedir = os.path.dirname(self.filename)
-        if not os.path.isdir(filedir):
-            os.makedirs(filedir)
-        self.filehandle = open(self.filename, "w")
-        self._write_header()
-
-    def _write_header(self):
-        self._get_header();
-        self.filehandle.write(ContentFile.header)
-
-    header=None
-
-    @classmethod
-    def _get_header(self):
-        """
-        Internal method to return the header for Contents.gz files
-
-        This is boilerplate which explains the contents of the file and how
-        it can be used.
-        """
-        if not ContentFile.header:
-            if Config().has_key("Contents::Header"):
-                try:
-                    h = open(os.path.join( Config()["Dir::Templates"],
-                                           Config()["Contents::Header"] ), "r")
-                    ContentFile.header = h.read()
-                    h.close()
-                except:
-                    log.error( "error opening header file: %d\n%s" % (Config()["Contents::Header"],
-                                                                      traceback.format_exc() ))
-                    ContentFile.header = None
-            else:
-                ContentFile.header = None
-
-        return ContentFile.header
-
-
-class DebContentFile(ContentFile):
-    def __init__(self,
-                 filename,
-                 suite_str,
-                 suite_id,
-                 arch_str,
-                 arch_id):
-        ContentFile.__init__(self,
-                             filename,
-                             suite_str,
-                             suite_id )
-        self.arch_str = arch_str
-        self.arch_id = arch_id
-
-    def query(self):
-        self.session = DBConn().session();
-
-        self.results = self.session.execute("""SELECT filename, comma_separated_list(section || '/' || package)
-        FROM deb_contents
-        WHERE ( arch=2 or arch = :arch) AND suite = :suite
-        """, { 'arch':self.arch_id, 'suite':self.suite_id } )
-
-class UdebContentFile(ContentFile):
-    def __init__(self,
-                 filename,
-                 suite_str,
-                 suite_id,
-                 section_name,
-                 section_id):
-        ContentFile.__init__(self,
-                             filename,
-                             suite_str,
-                             suite_id )
-
-    def query(self):
-        self.session = DBConn().session();
-
-        self.results = self.session.execute("""SELECT filename, comma_separated_list(section || '/' || package)
-        FROM udeb_contents
-        WHERE suite = :suite
-        group by filename
-        """ , { 'suite': self.suite_id } )
-
-class Contents(object):
-    """
-    Class capable of generating Contents-$arch.gz files
-    """
-    def __init__(self):
-        self.header = None
-
-    def reject(self, message):
-        log.error("E: %s" % message)
-
-    def cruft(self):
-        """
-        remove files/paths from the DB which are no longer referenced
-        by binaries and clean the temporary table
-        """
-        s = DBConn().session()
-
-        # clear out all of the temporarily stored content associations
-        # this should be run only after p-a has run.  after a p-a
-        # run we should have either accepted or rejected every package
-        # so there should no longer be anything in the queue
-        s.query(PendingContentAssociation).delete()
-
-        # delete any filenames we are storing which have no binary associated
-        # with them
-        cafq = s.query(ContentAssociation.filename_id).distinct()
-        cfq = s.query(ContentFilename)
-        cfq = cfq.filter(~ContentFilename.cafilename_id.in_(cafq))
-        cfq.delete()
-
-        # delete any paths we are storing which have no binary associated with
-        # them
-        capq = s.query(ContentAssociation.filepath_id).distinct()
-        cpq = s.query(ContentFilepath)
-        cpq = cpq.filter(~ContentFilepath.cafilepath_id.in_(capq))
-        cpq.delete()
-
-        s.commit()
-
-
-    def bootstrap_bin(self):
-        """
-        scan the existing debs in the pool to populate the bin_contents table
-        """
-        pooldir = Config()[ 'Dir::Pool' ]
-
-        s = DBConn().session()
-
-        for binary in s.query(DBBinary).yield_per(100):
-            print( "binary: %s" % binary.package )
-            filename = binary.poolfile.filename
-             # Check for existing contents
-            existingq = s.execute( "select 1 from bin_contents where binary_id=:id", {'id':binary.binary_id} );
-            if existingq.fetchone():
-                log.debug( "already imported: %s" % (filename))
-            else:
-                # We don't have existing contents so import them
-                log.debug( "scanning: %s" % (filename) )
-
-                debfile = os.path.join(pooldir, filename)
-                if os.path.exists(debfile):
-                    Binary(debfile, self.reject).scan_package(binary.binary_id, True)
-                else:
-                    log.error("missing .deb: %s" % filename)
-
-
-
-    def bootstrap(self):
-        """
-        scan the existing debs in the pool to populate the contents database tables
-        """
-        s = DBConn().session()
-
-
-        # get a mapping of all the override types we care about (right now .deb an .udeb)
-        override_type_map = {};
-        for override_type in s.query(OverrideType).all():
-            if override_type.overridetype.endswith('deb' ):
-                override_type_map[override_type.overridetype_id] = override_type.overridetype;
-
-        for override in s.query(Override).yield_per(100):
-            if not override_type_map.has_key(override.overridetype_id):
-                #this isn't an override we care about
-                continue
-
-            binaries = s.execute("""SELECT b.id, b.architecture
-                                    FROM binaries b
-                                    JOIN bin_associations ba ON ba.bin=b.id
-                                    WHERE ba.suite=:suite
-                                    AND b.package=:package""", {'suite':override.suite_id, 'package':override.package})
-            while True:
-                binary = binaries.fetchone()
-                if not binary:
-                    break
-
-                exists = s.execute("SELECT 1 FROM %s_contents WHERE binary_id=:id limit 1" % override_type_map[override.overridetype_id], {'id':binary.id})
-
-
-                if exists.fetchone():
-                    print '.',
-                    continue
-                else:
-                    print '+',
-
-                s.execute( """INSERT INTO %s_contents (filename,section,package,binary_id,arch,suite)
-                              SELECT file, :section, :package, :binary_id, :arch, :suite
-                              FROM bin_contents
-                              WHERE binary_id=:binary_id;""" % override_type_map[override.overridetype_id],
-                           { 'section' : override.section_id,
-                             'package' : override.package,
-                             'binary_id' : binary.id,
-                             'arch' : binary.architecture,
-                             'suite' : override.suite_id } )
-                s.commit()
-
-    def generate(self):
-        """
-        Generate contents files for both deb and udeb
-        """
-        self.deb_generate()
-        self.udeb_generate()
-
-    def deb_generate(self):
-        """
-        Generate Contents-$arch.gz files for every available arch in each given suite.
-        """
-        session = DBConn().session()
-        debtype_id = get_override_type("deb", session)
-        suites = self._suites()
-
-        inputtoquery = OneAtATime()
-        querytoingest = OneAtATime()
-        ingesttosort = OneAtATime()
-        sorttooutput = OneAtATime()
-        outputtogzip = OneAtATime()
-
-        qt = QueryThread(inputtoquery,querytoingest)
-        it = IngestThread(querytoingest,ingesttosort)
-        st = SortThread(ingesttosort,sorttooutput)
-        ot = OutputThread(sorttooutput,outputtogzip)
-        gt = GzipThread(outputtogzip, None)
-
-        qt.start()
-        it.start()
-        st.start()
-        ot.start()
-        gt.start()
-
-        # Get our suites, and the architectures
-        for suite in [i.lower() for i in suites]:
-            suite_id = get_suite(suite, session).suite_id
-            print( "got suite_id: %s for suite: %s" % (suite_id, suite ) )
-            arch_list = self._arches(suite_id, session)
-
-            for (arch_id,arch_str) in arch_list:
-                print( "suite: %s, arch: %s time: %s" %(suite_id, arch_id, datetime.datetime.now().isoformat()) )
-
-                filename = "dists/%s/Contents-%s" % (suite, arch_str)
-                cf = DebContentFile(filename, suite, suite_id, arch_str, arch_id)
-                inputtoquery.enqueue( cf )
-
-        inputtoquery.enqueue( EndOfContents() )
-        gt.join()
-
-    def udeb_generate(self):
-        """
-        Generate Contents-$arch.gz files for every available arch in each given suite.
-        """
-        session = DBConn().session()
-        udebtype_id=DBConn().get_override_type_id("udeb")
-        suites = self._suites()
-
-        inputtoquery = OneAtATime()
-        querytoingest = OneAtATime()
-        ingesttosort = OneAtATime()
-        sorttooutput = OneAtATime()
-        outputtogzip = OneAtATime()
-
-        qt = QueryThread(inputtoquery,querytoingest)
-        it = IngestThread(querytoingest,ingesttosort)
-        st = SortThread(ingesttosort,sorttooutput)
-        ot = OutputThread(sorttooutput,outputtogzip)
-        gt = GzipThread(outputtogzip, None)
-
-        qt.start()
-        it.start()
-        st.start()
-        ot.start()
-        gt.start()
-
-
-    def generate(self):
-        """
-        Generate Contents-$arch.gz files for every available arch in each given suite.
-        """
-        session = DBConn().session()
-
-        arch_all_id = get_architecture("all", session).arch_id
-
-        # The MORE fun part. Ok, udebs need their own contents files, udeb, and udeb-nf (not-free)
-        # This is HORRIBLY debian specific :-/
-        for dtype, section, fn_pattern in \
-              [('deb',  None,                        "dists/%s/Contents-%s.gz"),
-               ('udeb', "debian-installer",          "dists/%s/Contents-udeb-%s.gz"),
-               ('udeb', "non-free/debian-installer", "dists/%s/Contents-udeb-nf-%s.gz")]:
-
-            overridetype = get_override_type(dtype, session)
-
-            # For udebs, we only look in certain sections (see the for loop above)
-            if section is not None:
-                section = get_section(section, session)
-
-            # Get our suites
-            for suite in which_suites(session):
-                # Which architectures do we need to work on
-                arch_list = get_suite_architectures(suite.suite_name, skipsrc=True, skipall=True, session=session)
-
-                # Set up our file writer dictionary
-                file_writers = {}
-                try:
-                    # One file writer per arch
-                    for arch in arch_list:
-                        file_writers[arch.arch_id] = GzippedContentWriter(fn_pattern % (suite, arch.arch_string))
-
-                    for r in get_suite_contents(suite, overridetype, section, session=session).fetchall():
-                        filename, section, package, arch_id = r
-
-                        if arch_id == arch_all_id:
-                            # It's arch all, so all contents files get it
-                            for writer in file_writers.values():
-                                writer.write(filename, section, package)
-                        else:
-                            if file_writers.has_key(arch_id):
-                                file_writers[arch_id].write(filename, section, package)
-
-                finally:
-                    # close all the files
-                    for writer in file_writers.values():
-                        writer.finish()
-    def _suites(self):
-        """
-        return a list of suites to operate on
-        """
-        if Config().has_key( "%s::%s" %(options_prefix,"Suite")):
-            suites = utils.split_args(Config()[ "%s::%s" %(options_prefix,"Suite")])
-        else:
-            suites = Config().SubTree("Suite").List()
-
-        return suites
-
-    def _arches(self, suite, session):
-        """
-        return a list of archs to operate on
-        """
-        arch_list = []
-        arches = session.execute(
-            """SELECT s.architecture, a.arch_string
-            FROM suite_architectures s
-            JOIN architecture a ON (s.architecture=a.id)
-            WHERE suite = :suite_id""",
-            {'suite_id':suite } )
-
-        while True:
-            r = arches.fetchone()
-            if not r:
-                break
-
-            if r[1] != "source" and r[1] != "all":
-                arch_list.append((r[0], r[1]))
-
-        return arch_list
-
-
-def scan_all(limit):
+def scan_all(cnf, limit):
+    Logger = daklog.Logger(cnf.Cnf, 'contents scan')
     result = ContentsScanner.scan_all(limit)
     processed = '%(processed)d packages processed' % result
     remaining = '%(remaining)d packages remaining' % result
     Logger.log([processed, remaining])
+    Logger.close()
 
 ################################################################################
 
 def main():
     cnf = Config()
-
-    arguments = [('h',"help", "%s::%s" % (options_prefix,"Help")),
-                 ('s',"suite", "%s::%s" % (options_prefix,"Suite"),"HasArg"),
-                 ('l',"limit", "%s::%s" % (options_prefix,"Limit"),"HasArg"),
-                 ('q',"quiet", "%s::%s" % (options_prefix,"Quiet")),
-                 ('v',"verbose", "%s::%s" % (options_prefix,"Verbose")),
+    cnf['Contents::Options::Help'] = ''
+    cnf['Contents::Options::Suite'] = ''
+    cnf['Contents::Options::Limit'] = ''
+    cnf['Contents::Options::Force'] = ''
+    arguments = [('h', "help",  'Contents::Options::Help'),
+                 ('s', "suite", 'Contents::Options::Suite', "HasArg"),
+                 ('l', "limit", 'Contents::Options::Limit', "HasArg"),
+                 ('f', "force", 'Contents::Options::Force'),
                 ]
+    args = apt_pkg.ParseCommandLine(cnf.Cnf, arguments, sys.argv)
+    options = cnf.SubTree('Contents::Options')
 
-    commands = {'generate' : Contents.generate,
-                'bootstrap_bin' : Contents.bootstrap_bin,
-                'bootstrap' : Contents.bootstrap,
-                'cruft' : Contents.cruft,
-                }
-
-    args = apt_pkg.ParseCommandLine(cnf.Cnf, arguments,sys.argv)
-
-    if (len(args) < 1) or not commands.has_key(args[0]):
+    if (len(args) != 1) or options['Help']:
         usage()
 
-    if cnf.has_key("%s::%s" % (options_prefix,"Help")):
-        usage()
-
-    level=logging.INFO
-    if cnf.has_key("%s::%s" % (options_prefix,"Quiet")):
-        level=logging.ERROR
-
-    elif cnf.has_key("%s::%s" % (options_prefix,"Verbose")):
-        level=logging.DEBUG
-
-
-    logging.basicConfig( level=level,
-                         format='%(asctime)s %(levelname)s %(message)s',
-                         stream = sys.stderr )
-
-    global Logger
-    Logger = daklog.Logger(cnf.Cnf, 'contents')
-
     limit = None
-    if cnf.has_key("%s::%s" % (options_prefix,"Limit")):
-        limit = cnf["%s::%s" % (options_prefix,"Limit")]
+    if len(options['Limit']) > 0:
+        limit = int(options['Limit'])
 
-    if args[0] == 'bootstrap_bin':
-        scan_all(limit)
-    else:
-        commands[args[0]](Contents())
+    if args[0] == 'scan':
+        scan_all(cnf, limit)
+        return
 
-    Logger.close()
+    suite_names = utils.split_args(options['Suite'])
+
+    force = bool(options['Force'])
 
-def which_suites(session):
-    """
-    return a list of suites to operate on
-    """
-    if Config().has_key( "%s::%s" %(options_prefix,"Suite")):
-        suites = utils.split_args(Config()[ "%s::%s" %(options_prefix,"Suite")])
-    else:
-        suites = Config().SubTree("Suite").List()
+    if args[0] == 'generate':
+        write_all(cnf, suite_names, force)
+        return
 
-    return [get_suite(s.lower(), session) for s in suites]
+    usage()
 
 
 if __name__ == '__main__':