Philipp Kern [Sun, 7 Sep 2008 21:46:51 +0000 (21:46 +0000)]
do not check for superfluous checksums-* entries for now
The current check for additional entries in checksums-* which are not
covered by entries in files is flawed because files could be modified
by autobyhand or sometimes when the file's already in the pool.
We need to rework this to look into the original files dict instead.
Philipp Kern [Sun, 7 Sep 2008 20:50:06 +0000 (22:50 +0200)]
dak/process_accepted.py (install): use dsc_file instead of file
* dak/process_accepted.py (install): better use dsc_file
instead of the (for the loop iteration) static file
variable
Signed-off-by: Philipp Kern <pkern@debian.org>
# Please enter the commit message for your changes.
# (Comment lines starting with '#' will not be included)
# On branch new-master
# Changes to be committed:
# (use "git reset HEAD <file>..." to unstage)
#
# modified: ChangeLog
# modified: dak/process_accepted.py
#
# Changed but not updated:
# (use "git add <file>..." to update what will be committed)
#
# modified: daklib/database.py
#
# Untracked files:
# (use "git add <file>..." to include in what will be committed)
#
# config/asterix/
# diff
# ftp/
# incoming/
# lock/
# log/
# queue/
# web/removals.txt
* commit 'pkern/master':
Revert "revert all my stupid commits, we'll try this again later when we have a test server"
Revert "and revert the remainder"
implement sensible handling of checksum fields in .changes and .dsc
Philipp Kern [Sat, 6 Sep 2008 11:35:42 +0000 (11:35 +0000)]
dak/process_new.py (is_source_in_queue_dir): access u instead of Upload
2008-09-06 Philipp Kern <pkern@debian.org>
* dak/process_new.py (is_source_in_queue_dir): Access the right
variable to check if the given entry in the queue is the sourceful
upload we are looking for.
Philipp Kern [Sun, 31 Aug 2008 09:37:33 +0000 (09:37 +0000)]
cosmetic fixes for do_accept and do_accept_stableupdate
2008-08-31 Philipp Kern <pkern@debian.org>
* dak/process_new.py (_accept): new helper function to accept
an upload regularly, obeying no-action if set
* dak/process_new.py (do_accept): use _accept
* dak/process_new.py (do_accept_stableupdate): bail out in else
on binary uploads, in case we missed something; use the _accept
helper; reordered the queue checks to do frequent cases first
Joerg Jaspert [Fri, 29 Aug 2008 20:25:10 +0000 (22:25 +0200)]
Merge commit 'pkern/master'
* commit 'pkern/master':
fix proposed-updates processing in dinstall: actually accept sourceful uploads
dak/process_new.py: do not try to free the unchecked lockfile in no-action mode
Philipp Kern [Thu, 28 Aug 2008 20:47:22 +0000 (22:47 +0200)]
implement sensible handling of checksum fields in .changes and .dsc
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
Philipp Kern [Tue, 26 Aug 2008 10:25:39 +0000 (10:25 +0000)]
fix proposed-updates processing in dinstall: actually accept sourceful uploads
* dak/process_new.py (is_source_in_queue_dir): fix variable usage
* dak/process_new.py (move_to_holding): just state what we intend
to do in no-action mode
* dak/process_new.py (do_accept_stableupdate): fetch summaries,
fix invokation of is_source_in_queue_dir, actually accept sourceful
uploads in p-u holding
Joerg Jaspert [Sat, 16 Aug 2008 05:15:56 +0000 (07:15 +0200)]
Merge commit 'mhy/checksums'
* commit 'mhy/checksums':
and revert the remainder
revert all my stupid commits, we'll try this again later when we have a test server
Revert "use correct field names"
use correct field names
Joerg Jaspert [Sat, 16 Aug 2008 02:15:16 +0000 (04:15 +0200)]
Merge commit 'mhy/checksums'
* commit 'mhy/checksums':
expand check_archive.py to cope with extra checksums
add changelog to keep Ganneff happy
add sha1sum and sha256sum columns
fix typos
make process accepted not die with old .dak files
an attempt to move the functionaliity into a module
let's store the known_hashes information centrally
first attempt at bodging in support for sha1sum and sha256sum
Joerg Jaspert [Tue, 12 Aug 2008 14:12:34 +0000 (16:12 +0200)]
* config/debian/cron.dinstall: Complicate the i18n export a little
by using date/hour based directories which we then link into the
web view. They contain a signed timestamp file now, which means
the i18n people can take a long time to generate files, yet we
still know exactly on which dataset their data is based on, and
can then verify it with that. Ensures we only get descriptions for
packages we know off (or knew of in the past 2 days).
Joerg Jaspert [Fri, 8 Aug 2008 18:50:12 +0000 (20:50 +0200)]
Merge commit 'pkern/master'
* commit 'pkern/master':
dak/process_new.py: fix up binary diversion into various queues
dak/process_new.py: handle uploads to (oldstable)-proposed-updates differently
Joerg Jaspert [Fri, 8 Aug 2008 12:24:41 +0000 (12:24 +0000)]
Gah, apt is a piece of shit. Well, it expects the uncompressed
Packages files in the Release file to appear, so lets generate them.
Also make sure we dont have empty release files listed...
Philipp Kern [Thu, 7 Aug 2008 22:57:11 +0000 (19:57 -0300)]
dak/process_new.py: handle uploads to (oldstable)-proposed-updates differently
Uploads targetting proposed-updates used to hit NEW first and were going
to proposed-updates without further review by SRM afterwards. This patch
fixes this behaviour by sending the upload to p-u-new after NEW processing.
If an upload is sourceful it is moved right away. If it is not its
corresponding source is searched in p-u holding, new, accepted and
in the archive and moved appropriately into the right queue.
Add a script to fetch the wanna-build-database from
http://buildd.debian.org/stats/, call this script from cron.weekly, and
document this change in the ChangeLog.
projects / dak / log