Patch from Nikita V. Youshchenko
From: "Nikita V. Youshchenko"
To: Debian Bug Tracking System
Subject: offlineimap: exception on mail with broken headers (+fix)
Date: Wed, 24 Aug 2005 13:41:08 +0400
Package: offlineimap
Version: 4.0.10
Severity: normal
Tags: patch
Recently I've got an exception (see below) while using offlineimap.
Exception was probably caused by invalid Date header of (likely spam)
message:
Date: Sat, 20 Aug 2005 4294967295:43:18 -0700
From: "Nikita V. Youshchenko"
I trued to use offlineimap and found that while being quite fast on
small folders, it takes up to several minutes (of 100% busy CPU and
almost no network traffic) to sync a folder with 2000+ messages.
While looking into the code, I found why this happens.
In folder/Base.py, in method BaseFolder.syncmessagesto_copy(),
dest.getmessagelist() is called inside a loop, while being a loop
invariant. Similar thing happens in BaseFolder.syncmessagesto_delete()
for self.getmessagelist().
This causes quadratic complexity over folder size.
Moving these calls out of loops make large folder sync fast (several
seconds instead of several minutes for folder with 2000 messages on
700MHz P3).
Applied patch from Joerg Wendland <joergland@debian.org> to use
APPENDUID result from mail servers that provide it. Closes: #198772.
Resolves: [debian.org #198772]
Added a "force" option to imapserver/select to force a reloading of a
folder. Per [complete.org #67], when cachemessagelist() was called on
an object that was cached from a previous run, it would not re-issue
the select(). Closes: [complete.org #67]
Added -l option. Updated documentation for it. Changed _msg to
_display override in UI modules. Renamed "doc" to "docs" target in
Makefile to avoid conflicting with a subdir.
- offlineimap (3.99.17) unstable; urgency=low
- Fixed two potential obscure race conditions in folder/Maildir.py. +
Condition 1 involved the gettimeseq() function. This function
accesses per-module variables but does not have a lock. It may have
been possible for this to have been called in such a way that
timeseq was not properly updated. + Condition 2 involved the call to
gettimeseq(). Since the timeseq is based on the system clock, we now
use the time as reported inside timeseq() rather than outside. This
way, we can be assured that the same value is in use both places.
- Added debug code to savemessage in folder/Maildir.py to try to track
down a mysterious 0-length file bug. -- John Goerzen
<jgoerzen@complete.org> Tue, 6 May 2003 09:21:38 -0500
- Added some significant debug code to folder/IMAP.py when saving a
new message with APPEND. This should make it easier to track down
bugs both in OfflineIMAP and in mail servers that implement this
poorly.
- Fixed adding of X-OfflineIMAP header when the message starts out
with no headers. (This should not generally occur.) This should help
with some "invalid literal for long()" problems.
When sep was /, the new Maildir support code would recursively try to
scan ., resulting in huge paths and an eventual crash. Fixed with a
one-line patch to Maildir.py. Closes: [complete.org #60] Sergei, The
below diff is going into 3.99.16. You can apply it to 3.99.15 and it
should work for you now. Please let me know. (Ignore any patch errors
for debian/changelog). Thanks for the report.
Raise an exception when the status area is locked. This will cause UIs
to go through their normal exception handling code. In particular, for
the Curses.Blinkenlights interface, the Curses module will be stopped
and the error message will be printed on the console. Previously, this
error message would not have been visible. Closes: #185709.
- Slight renaming in offlineimap.conf.minimal to clarify things.
- Documentation updated with information about new features. Closes:
#189771. + Described IMAP-IMAP syncing + Updated minimal example
with new offlineimap.conf.minimal + Updated UID information. Added
link to recent mailing list discussion. + Described KMail syncing,
which now works. + Added link to mailing list archives.
- Now checks that SELECT succeeded when entering a folder.
- Verifies that folders listed on folderincludes actually exist by
trying to enter them. Thus, if they do not exist, they can be
created on the first run.
Fixed line-ending code to deal with files with mixed \n and \r\n
codes. This is a rare case, but now is more onerous because we now
have to find headers.
Due to possibly having one account sleep while another is reading a
password, and other tricky situations, support for nice updating and
cancelling of a sleep in TTY.TTYUI has been removed. However, this is
not going to be a huge problem because the new Curses Blinkenlights
interface has this support, and does it a lot better than TTY.TTYUI
ever could have.
Reworked the canvas. Before, problem was the label and buttons to the
right of the lights would make the window too wide. When the button
got added, the window would get even wider. That was because the
canvas would not shrink. My workaround is to use a separate canvas for
each light. Seems to be OK here....
Fixed up the VerboseUI for new account system. All that really needed
updating with the "Sync immediately" button, to cope with syncing
different accounts at different times. It's better now.
More locking updates. Introduced a new MultiLock to threadutil. This
lock will let a single thread acquire the same lock more than once,
keeping track of how many times this happens, and will release the
actual lock only when the lock's lock count gets back to zero. By
using MultiLock, various functions in Curses.py and Blinkenlights.py
no longer need to pass around to other functions a parameter
indicating whether or not a lock should be obtained. This was a large
cause of complexity and errors, which is now eliminated. Everything
seems to be working properly wrt locking at this point. The
Curses.Blinkenlights interface has achieved basic working
functionality.
Updated the mbnames recorder to bring it back up-to-date with the new
account-centric system. It will now gather reports from account sync
threads, and when it has all that it's supposed to, it'll write out
the file.
Added some temporary debug code to help weed out a few race conditions
with the curses Blinkenlights interface. Think I've finally got it.
I'm leaving the debugging code in for now, though, to help in case
there are future problems.
More progress at debugging. The curses blinkenlights is now working
well, though it still has an occasional tendency to corrupt the light
display with comments from the log. I suspect a locking problem --
need to be more strict with iolock I suspect. Updated various modules
to register the threads' account names, etc.