+2008-08-28 Philipp Kern <pkern@debian.org>
+
+ * daklib/utils.py (check_hashes): adapt to different API, check
+ sizes separately
+
+ * daklib/utils.py (parse_changes, parse_deb822): refactor
+ the string-based logic of parse_changes into a new function
+ parse_deb822; parse_changes itself remains file-based
+
+ * daklib/utils.py (hash_key): gives the key of a hash in the
+ files dict
+
+ * daklib/utils.py (create_hash, check_size): made more readable
+
+ * daklib/utils.py (check_hash): just check the hashes and complain
+ about missing checksums
+
+ * daklib/utils.py (check_hash_fields): function to reject unknown
+ checksums fields
+
+ * daklib/utils.py (_ensure_changes_hash, _ensure_dsc_hash): helper
+ functions for ensure_hashes; check their corresponding manifests'
+ hashes
+
+ * daklib/utils.py (ensure_hashes): retrieve the checksums fields
+ from the original filecontents blob so that they do not need to
+ be present in the .dak; refactored the actual checks by calling
+ the aforementioned helper functions
+
+ * daklib/utils.py (parse_checksums): parse a given checksums field
+ in a manifest and insert the values found into the files dict,
+ checking the file sizes on the way
+
2008-09-06 Philipp Kern <pkern@debian.org>
* dak/process_new.py (is_source_in_queue_dir): Access the right
* config/debian/cron.dinstall: We dont want i18n to ever fail
dinstall, add a || true
+2008-08-15 Mark Hymers <mhy@debian.org>
+
+ * daklib/utils.py: Actually import a module before using it.
+
+ * daklib/utils.py: Actually check we have basedict before trying to
+ use it.
+
+ * dak/process_accepted.py, dak/process_unchecked.py,
+ daklib/database.py: Don't change get_files_id to use sha1sum and
+ sha256sum.
+
+ * setup/init_pool.sql, dak/check_archive.py, dak/decode_dot_dak.py,
+ dak/process_accepted.py, dak/process_unchecked.py, daklib/database.py,
+ daklib/queue.py, daklib/utils.py: Attempt to add sha1sum and
+ sha256sums into the database. The complication is that we have to
+ keep backwards compatibility with the .dak files already in existance.
+ Note that import_archive hasn't been hacked to deal with this yet.
+
2008-08-14 Joerg Jaspert <joerg@debian.org>
* config/debian/cron.dinstall: Added the i18n retrieval of package
2008-08-07 Stephen Gran <sgran@debian.org>
* Drop use of exec to eval variable interpolation
-
2008-08-07 Joerg Jaspert <joerg@debian.org>
* dak/process_accepted.py (install): Error out with the new
The following MODEs are available:
- md5sums - validate the md5sums stored in the database
+ checksums - validate the checksums stored in the database
files - check files in the database against what's in the archive
dsc-syntax - validate the syntax of .dsc files in the archive
missing-overrides - check for missing overrides
################################################################################
-def check_md5sums():
+def check_checksums():
print "Getting file information from database..."
- q = projectB.query("SELECT l.path, f.filename, f.md5sum, f.size FROM files f, location l WHERE f.location = l.id")
+ q = projectB.query("SELECT l.path, f.filename, f.md5sum, f.sha1sum, f.sha256sum, f.size FROM files f, location l WHERE f.location = l.id")
ql = q.getresult()
- print "Checking file md5sums & sizes..."
+ print "Checking file checksums & sizes..."
for i in ql:
filename = os.path.abspath(i[0] + i[1])
db_md5sum = i[2]
- db_size = int(i[3])
+ db_sha1sum = i[3]
+ db_sha256sum = i[4]
+ db_size = int(i[5])
try:
f = utils.open_file(filename)
except:
utils.warn("**WARNING** md5sum mismatch for '%s' ('%s' [current] vs. '%s' [db])." % (filename, md5sum, db_md5sum))
if size != db_size:
utils.warn("**WARNING** size mismatch for '%s' ('%s' [current] vs. '%s' [db])." % (filename, size, db_size))
+ # Until the main database is filled, we need to not spit 500,000 warnings
+ # every time we scan the archive. Yet another hack (TM) which can go away
+ # once this is all working
+ if db_sha1sum is not None and db_sha1sum != '':
+ sha1sum = apt_pkg.sha1sum(f)
+ if sha1sum != db_sha1sum:
+ utils.warn("**WARNING** sha1sum mismatch for '%s' ('%s' [current] vs. '%s' [db])." % (filename, sha1sum, db_sha1sum))
+
+ if db_sha256sum is not None and db_sha256sum != '':
+ sha256sum = apt_pkg.sha256sum(f)
+ if sha256sum != db_sha256sum:
+ utils.warn("**WARNING** sha256sum mismatch for '%s' ('%s' [current] vs. '%s' [db])." % (filename, sha256sum, db_sha256sum))
print "Done."
projectB = pg.connect(Cnf["DB::Name"], Cnf["DB::Host"], int(Cnf["DB::Port"]))
database.init(Cnf, projectB)
- if mode == "md5sums":
- check_md5sums()
+ if mode == "checksums":
+ check_checksums()
elif mode == "files":
check_files()
elif mode == "dsc-syntax":
for f in files.keys():
print " %s:" % (f)
for i in [ "package", "version", "architecture", "type", "size",
- "md5sum", "component", "location id", "source package",
- "source version", "maintainer", "dbtype", "files id",
- "new", "section", "priority", "pool name" ]:
+ "md5sum", "sha1sum", "sha256sum", "component", "location id",
+ "source package", "source version", "maintainer", "dbtype",
+ "files id", "new", "section", "priority", "pool name" ]:
if files[f].has_key(i):
print " %s: %s" % (i.capitalize(), files[f][i])
del files[f][i]
# Begin a transaction; if we bomb out anywhere between here and the COMMIT WORK below, the DB will not be changed.
projectB.query("BEGIN WORK")
+ # Check the hashes are all present: HACK: Can go away once all dak files
+ # are known to be newer than the shasum changes
+ utils.ensure_hashes(changes, dsc, files, dsc_files)
+
# Add the .dsc file to the DB
for file in files.keys():
if files[file]["type"] == "dsc":
dsc_component = files[file]["component"]
dsc_location_id = files[file]["location id"]
if not files[file].has_key("files id") or not files[file]["files id"]:
- files[file]["files id"] = database.set_files_id (filename, files[file]["size"], files[file]["md5sum"], dsc_location_id)
+ files[file]["files id"] = database.set_files_id (filename, files[file]["size"], files[file]["md5sum"], files[file]["sha1sum"], files[file]["sha256sum"], dsc_location_id)
projectB.query("INSERT INTO source (source, version, maintainer, changedby, file, install_date, sig_fpr) VALUES ('%s', '%s', %d, %d, %d, '%s', %s)"
% (package, version, maintainer_id, changedby_id, files[file]["files id"], install_date, fingerprint_id))
files_id = database.get_files_id(filename, dsc_files[dsc_file]["size"], dsc_files[dsc_file]["md5sum"], dsc_location_id)
# FIXME: needs to check for -1/-2 and or handle exception
if files_id == None:
- files_id = database.set_files_id (filename, dsc_files[dsc_file]["size"], dsc_files[dsc_file]["md5sum"], dsc_location_id)
+ files_id = database.set_files_id (filename, dsc_files[dsc_file]["size"], dsc_files[dsc_file]["md5sum"], files[file]["sha1sum"], files[file]["sha256sum"], dsc_location_id)
projectB.query("INSERT INTO dsc_files (source, file) VALUES (currval('source_id_seq'), %d)" % (files_id))
# Add the src_uploaders to the DB
if not files[file].has_key("location id") or not files[file]["location id"]:
files[file]["location id"] = database.get_location_id(Cnf["Dir::Pool"],files[file]["component"],utils.where_am_i())
if not files[file].has_key("files id") or not files[file]["files id"]:
- files[file]["files id"] = database.set_files_id (filename, files[file]["size"], files[file]["md5sum"], files[file]["location id"])
+ files[file]["files id"] = database.set_files_id (filename, files[file]["size"], files[file]["md5sum"], files[file]["sha1sum"], files[file]["sha256sum"], files[file]["location id"])
source_id = database.get_source_id (source, source_version)
if source_id:
projectB.query("INSERT INTO binaries (package, version, maintainer, source, architecture, file, type, sig_fpr) VALUES ('%s', '%s', %d, %d, %d, %d, '%s', %d)"
#
if changes["architecture"].has_key("source") and orig_tar_id and \
orig_tar_location != "legacy" and orig_tar_location != dsc_location_id:
- q = projectB.query("SELECT l.path, f.filename, f.size, f.md5sum FROM files f, location l WHERE f.id = %s AND f.location = l.id" % (orig_tar_id))
+ q = projectB.query("SELECT l.path, f.filename, f.size, f.md5sum, f.sha1sum, f.sha256sum FROM files f, location l WHERE f.id = %s AND f.location = l.id" % (orig_tar_id))
ql = q.getresult()[0]
old_filename = ql[0] + ql[1]
file_size = ql[2]
file_md5sum = ql[3]
+ file_sha1sum = ql[4]
+ file_sha256sum = ql[5]
new_filename = utils.poolify(changes["source"], dsc_component) + os.path.basename(old_filename)
new_files_id = database.get_files_id(new_filename, file_size, file_md5sum, dsc_location_id)
if new_files_id == None:
utils.copy(old_filename, Cnf["Dir::Pool"] + new_filename)
- new_files_id = database.set_files_id(new_filename, file_size, file_md5sum, dsc_location_id)
+ new_files_id = database.set_files_id(new_filename, file_size, file_md5sum, file_sha1sum, file_sha256sum, dsc_location_id)
projectB.query("UPDATE dsc_files SET file = %s WHERE source = %s AND file = %s" % (new_files_id, source_id, orig_tar_id))
# Install the files into the pool
files[orig_tar_gz] = {}
files[orig_tar_gz]["size"] = os.stat(orig_tar_gz)[stat.ST_SIZE]
files[orig_tar_gz]["md5sum"] = dsc_files[orig_tar_gz]["md5sum"]
+ files[orig_tar_gz]["sha1sum"] = dsc_files[orig_tar_gz]["sha1sum"]
+ files[orig_tar_gz]["sha256sum"] = dsc_files[orig_tar_gz]["sha256sum"]
files[orig_tar_gz]["section"] = files[dsc_filename]["section"]
files[orig_tar_gz]["priority"] = files[dsc_filename]["priority"]
files[orig_tar_gz]["component"] = files[dsc_filename]["component"]
################################################################################
def check_hashes ():
- # Make sure we recognise the format of the Files: field
- format = changes.get("format", "0.0").split(".",1)
- if len(format) == 2:
- format = int(format[0]), int(format[1])
- else:
- format = int(float(format[0])), 0
-
- check_hash(".changes", files, "md5sum", apt_pkg.md5sum)
- check_hash(".dsc", dsc_files, "md5sum", apt_pkg.md5sum)
-
- if format >= (1,8):
- hashes = [("sha1", apt_pkg.sha1sum),
- ("sha256", apt_pkg.sha256sum)]
- else:
- hashes = []
-
- for x in changes:
- if x.startswith("checksum-"):
- h = x.split("-",1)[1]
- if h not in dict(hashes):
- reject("Unsupported checksum field in .changes" % (h))
-
- for x in dsc:
- if x.startswith("checksum-"):
- h = x.split("-",1)[1]
- if h not in dict(hashes):
- reject("Unsupported checksum field in .dsc" % (h))
-
- for h,f in hashes:
- try:
- fs = utils.build_file_list(changes, 0, "checksums-%s" % h, h)
- check_hash(".changes %s" % (h), fs, h, f, files)
- except NoFilesFieldError:
- reject("No Checksums-%s: field in .changes" % (h))
- except UnknownFormatError, format:
- reject("%s: unknown format of .changes" % (format))
- except ParseChangesError, line:
- reject("parse error for Checksums-%s in .changes, can't grok: %s." % (h, line))
-
- if "source" not in changes["architecture"]: continue
-
- try:
- fs = utils.build_file_list(dsc, 1, "checksums-%s" % h, h)
- check_hash(".dsc %s" % (h), fs, h, f, dsc_files)
- except UnknownFormatError, format:
- reject("%s: unknown format of .dsc" % (format))
- except NoFilesFieldError:
- reject("No Checksums-%s: field in .dsc" % (h))
- except ParseChangesError, line:
- reject("parse error for Checksums-%s in .dsc, can't grok: %s." % (h, line))
-
-################################################################################
-
-def check_hash (where, lfiles, key, testfn, basedict = None):
- if basedict:
- for f in basedict.keys():
- if f not in lfiles:
- reject("%s: no %s checksum" % (f, key))
-
- for f in lfiles.keys():
- if basedict and f not in basedict:
- reject("%s: extraneous entry in %s checksums" % (f, key))
-
- try:
- file_handle = utils.open_file(f)
- except CantOpenError:
- continue
-
- # Check hash
- if testfn(file_handle) != lfiles[f][key]:
- reject("%s: %s check failed." % (f, key))
- file_handle.close()
- # Check size
- actual_size = os.stat(f)[stat.ST_SIZE]
- size = int(lfiles[f]["size"])
- if size != actual_size:
- reject("%s: actual file size (%s) does not match size (%s) in %s"
- % (f, actual_size, size, where))
+ utils.check_hash(".changes", files, "md5", apt_pkg.md5sum)
+ utils.check_size(".changes", files)
+ utils.check_hash(".dsc", dsc_files, "md5", apt_pkg.md5sum)
+ utils.check_size(".dsc", dsc_files)
+
+ # This is stupid API, but it'll have to do for now until
+ # we actually have proper abstraction
+ for m in utils.ensure_hashes(changes, dsc, files, dsc_files):
+ reject(m)
################################################################################
################################################################################
-def set_files_id (filename, size, md5sum, location_id):
+def set_files_id (filename, size, md5sum, sha1sum, sha256sum, location_id):
global files_id_cache
- projectB.query("INSERT INTO files (filename, size, md5sum, location) VALUES ('%s', %d, '%s', %d)" % (filename, long(size), md5sum, location_id))
+ projectB.query("INSERT INTO files (filename, size, md5sum, sha1sum, sha256sum, location) VALUES ('%s', %d, '%s', '%s', '%s', %d)" % (filename, long(size), md5sum, sha1sum, sha256sum, location_id))
return get_files_id (filename, size, md5sum, location_id)
for file_entry in files.keys():
d_files[file_entry] = {}
for i in [ "package", "version", "architecture", "type", "size",
- "md5sum", "component", "location id", "source package",
- "source version", "maintainer", "dbtype", "files id",
- "new", "section", "priority", "othercomponents",
+ "md5sum", "sha1sum", "sha256sum", "component",
+ "location id", "source package", "source version",
+ "maintainer", "dbtype", "files id", "new",
+ "section", "priority", "othercomponents",
"pool name", "original component" ]:
if files[file_entry].has_key(i):
d_files[file_entry][i] = files[file_entry][i]
#!/usr/bin/env python
+# vim:set et ts=4 sw=4:
# Utility functions
# Copyright (C) 2000, 2001, 2002, 2003, 2004, 2005, 2006 James Troup <james@nocrew.org>
################################################################################
import codecs, commands, email.Header, os, pwd, re, select, socket, shutil, \
- sys, tempfile, traceback
+ sys, tempfile, traceback, stat
import apt_pkg
import database
from dak_exceptions import *
alias_cache = None
key_uid_email_cache = {}
+# (hashname, function, earliest_changes_version)
+known_hashes = [("sha1", apt_pkg.sha1sum, (1, 8)),
+ ("sha256", apt_pkg.sha256sum, (1, 8))]
+
################################################################################
def open_file(filename, mode='r'):
################################################################################
-def parse_changes(filename, signing_rules=0):
- """Parses a changes file and returns a dictionary where each field is a
-key. The mandatory first argument is the filename of the .changes
-file.
-
-signing_rules is an optional argument:
-
- o If signing_rules == -1, no signature is required.
- o If signing_rules == 0 (the default), a signature is required.
- o If signing_rules == 1, it turns on the same strict format checking
- as dpkg-source.
-
-The rules for (signing_rules == 1)-mode are:
-
- o The PGP header consists of "-----BEGIN PGP SIGNED MESSAGE-----"
- followed by any PGP header data and must end with a blank line.
-
- o The data section must end with a blank line and must be followed by
- "-----BEGIN PGP SIGNATURE-----".
-"""
-
+def parse_deb822(contents, signing_rules=0):
error = ""
changes = {}
- changes_in = open_file(filename)
- lines = changes_in.readlines()
+ # Split the lines in the input, keeping the linebreaks.
+ lines = contents.splitlines(True)
- if not lines:
+ if len(lines) == 0:
raise ParseChangesError, "[Empty changes file]"
# Reindex by line number so we can easily verify the format of
if signing_rules == 1 and inside_signature:
raise InvalidDscError, index
- changes_in.close()
changes["filecontents"] = "".join(lines)
if changes.has_key("source"):
################################################################################
+def parse_changes(filename, signing_rules=0):
+ """Parses a changes file and returns a dictionary where each field is a
+key. The mandatory first argument is the filename of the .changes
+file.
+
+signing_rules is an optional argument:
+
+ o If signing_rules == -1, no signature is required.
+ o If signing_rules == 0 (the default), a signature is required.
+ o If signing_rules == 1, it turns on the same strict format checking
+ as dpkg-source.
+
+The rules for (signing_rules == 1)-mode are:
+
+ o The PGP header consists of "-----BEGIN PGP SIGNED MESSAGE-----"
+ followed by any PGP header data and must end with a blank line.
+
+ o The data section must end with a blank line and must be followed by
+ "-----BEGIN PGP SIGNATURE-----".
+"""
+
+ changes_in = open_file(filename)
+ content = changes_in.read()
+ changes_in.close()
+ return parse_deb822(content, signing_rules)
+
+################################################################################
+
+def hash_key(hashname):
+ return '%ssum' % hashname
+
+################################################################################
+
+def create_hash(where, files, hashname, hashfunc):
+ """create_hash extends the passed files dict with the given hash by
+ iterating over all files on disk and passing them to the hashing
+ function given."""
+
+ rejmsg = []
+ for f in files.keys():
+ try:
+ file_handle = open_file(f)
+ except CantOpenError:
+ rejmsg.append("Could not open file %s for checksumming" % (f))
+
+ files[f][hash_key(hashname)] = hashfunc(file_handle)
+
+ file_handle.close()
+ return rejmsg
+
+################################################################################
+
+def check_hash(where, files, hashname, hashfunc):
+ """check_hash checks the given hash in the files dict against the actual
+ files on disk. The hash values need to be present consistently in
+ all file entries. It does not modify its input in any way."""
+
+ rejmsg = []
+ for f in files.keys():
+ try:
+ file_handle = open_file(f)
+
+ # Check for the hash entry, to not trigger a KeyError.
+ if not files[f].has_key(hash_key(hashname)):
+ rejmsg.append("%s: misses %s checksum in %s" % (f, hashname,
+ where))
+ continue
+
+ # Actually check the hash for correctness.
+ if hashfunc(file_handle) != files[f][hash_key(hashname)]:
+ rejmsg.append("%s: %s check failed in %s" % (f, hashname,
+ where))
+ except CantOpenError:
+ # XXX: IS THIS THE BLOODY CASE WHEN THE FILE'S IN THE POOL!?
+ continue
+ finally:
+ file_handle.close()
+ return rejmsg
+
+################################################################################
+
+def check_size(where, files):
+ """check_size checks the file sizes in the passed files dict against the
+ files on disk."""
+
+ rejmsg = []
+ for f in files.keys():
+ actual_size = os.stat(f)[stat.ST_SIZE]
+ size = int(files[f]["size"])
+ if size != actual_size:
+ rejmsg.append("%s: actual file size (%s) does not match size (%s) in %s"
+ % (f, actual_size, size, where))
+ return rejmsg
+
+################################################################################
+
+def check_hash_fields(what, manifest):
+ """check_hash_fields ensures that there are no checksum fields in the
+ given dict that we do not know about."""
+
+ rejmsg = []
+ hashes = map(lambda x: x[0], known_hashes)
+ for field in manifest:
+ if field.startswith("checksums-"):
+ hashname = field.split("-",1)[1]
+ if hashname not in hashes:
+ rejmsg.append("Unsupported checksum field for %s "\
+ "in %s" % (hashname, what))
+ return rejmsg
+
+################################################################################
+
+def _ensure_changes_hash(changes, format, version, files, hashname, hashfunc):
+ if format >= version:
+ # The version should contain the specified hash.
+ func = check_hash
+
+ # Import hashes from the changes
+ rejmsg = parse_checksums(".changes", files, changes, hashname)
+ if len(rejmsg) > 0:
+ return rejmsg
+ else:
+ # We need to calculate the hash because it can't possibly
+ # be in the file.
+ func = create_hash
+ return func(".changes", files, hashname, hashfunc)
+
+# We could add the orig which might be in the pool to the files dict to
+# access the checksums easily.
+
+def _ensure_dsc_hash(dsc, dsc_files, hashname, hashfunc):
+ """ensure_dsc_hashes' task is to ensure that each and every *present* hash
+ in the dsc is correct, i.e. identical to the changes file and if necessary
+ the pool. The latter task is delegated to check_hash."""
+
+ rejmsg = []
+ if not dsc.has_key('Checksums-%s' % (hashname,)):
+ return rejmsg
+ # Import hashes from the dsc
+ parse_checksums(".dsc", dsc_files, dsc, hashname)
+ # And check it...
+ rejmsg.extend(check_hash(".dsc", dsc_files, hashname, hashfunc))
+ return rejmsg
+
+################################################################################
+
+def ensure_hashes(changes, dsc, files, dsc_files):
+ rejmsg = []
+
+ # Make sure we recognise the format of the Files: field in the .changes
+ format = changes.get("format", "0.0").split(".", 1)
+ if len(format) == 2:
+ format = int(format[0]), int(format[1])
+ else:
+ format = int(float(format[0])), 0
+
+ # We need to deal with the original changes blob, as the fields we need
+ # might not be in the changes dict serialised into the .dak anymore.
+ orig_changes = parse_deb822(changes['filecontents'])
+
+ # Copy the checksums over to the current changes dict. This will keep
+ # the existing modifications to it intact.
+ for field in orig_changes:
+ if field.startswith('checksums-'):
+ changes[field] = orig_changes[field]
+
+ # Check for unsupported hashes
+ rejmsg.extend(check_hash_fields(".changes", changes))
+ rejmsg.extend(check_hash_fields(".dsc", dsc))
+
+ # We have to calculate the hash if we have an earlier changes version than
+ # the hash appears in rather than require it exist in the changes file
+ for hashname, hashfunc, version in known_hashes:
+ rejmsg.extend(_ensure_changes_hash(changes, format, version, files,
+ hashname, hashfunc))
+ if "source" in changes["architecture"]:
+ rejmsg.extend(_ensure_dsc_hash(dsc, dsc_files, hashname,
+ hashfunc))
+
+ return rejmsg
+
+def parse_checksums(where, files, manifest, hashname):
+ rejmsg = []
+ field = 'checksums-%s' % hashname
+ if not field in manifest:
+ return rejmsg
+ input = manifest[field]
+ for line in input.split('\n'):
+ if not line:
+ break
+ hash, size, file = line.strip().split(' ')
+ if not files.has_key(file):
+ rejmsg.append("%s: not present in files but in checksums-%s in %s" %
+ (file, hashname, where))
+ if not files[file]["size"] == size:
+ rejmsg.append("%s: size differs for files and checksums-%s entry "\
+ "in %s" % (file, hashname, where))
+ files[file][hash_key(hashname)] = hash
+ for f in files.keys():
+ if not files[f].has_key(hash_key(hashname)):
+ rejmsg.append("%s: no entry in checksums-%s in %s" % (file,
+ hashname, where))
+ return rejmsg
+
+################################################################################
+
# Dropped support for 1.4 and ``buggy dchanges 3.4'' (?!) compared to di.pl
def build_file_list(changes, is_a_dsc=0, field="files", hashname="md5sum"):
md5sum TEXT NOT NULL,
location INT4 NOT NULL, -- REFERENCES location
last_used TIMESTAMP,
+ sha1sum TEXT NOT NULL,
+ sha256sum TEXT NOT NULL,
unique (filename, location)
);