- We used to give a list of pointers of messages to upload to the server,
which is really bad, since the user could delete one before the job would
get around to upload it. Now it works on serial numbers instead and
there are no more crashes in this area
- We had two QDicts: IMAP UID -> KMail index and index -> UID. This was
not good because the maps would get out of sync and result in rebuilding
way too many times, plus the map from index to UID was not necessary.
Now the redundant map is gone, and a dirty flag was introduced to only
rebuild when necessary. This was one of the more serious reasons we
had dIMAP do 100% cpu time because it ended up in a very time consuming
O(n^2) remap method
- For stability reasons, we need to save the highest UID number we have
downloaded from the server fairly often. We took this seriously and
saved on every message. Another of the reasons it was incredibly slow.
Now it's timed, so there will only be a save a minute
- Don't write a uid cache file when we don't have any info for it
- Delete the uid cache file from the dimap cache directory
- Don't rebuild the remaining uid map all the time
- Fix bug 68587 - delete the relevant folders from the folder tree when
the account is removed
Patch reviewed by Till.
Bo.
svn path=/trunk/kdepim/; revision=276112
active readerwindow to ask that for the attachmentstrategy but pass it
along with the folder job, so it is available when the mime structure
comes in with load on demand. Remove unused KMKernel::activeReaderWindow
which doesn't work in kontact anyhow.
CCMAIL: 67644-done@bugs.kde.org
svn path=/trunk/kdepim/; revision=268877
signals from being emitted.
This should fix the sent folder count updating bug. If it does not for
you, please reopen.
Patch Ok'ed by Till (Don also agreed that quiet() can be removed).
CCMAIL: 63670-done@bugs.kde.org
svn path=/trunk/kdepim/; revision=260199
#define kernel KMKernel::self()
to
#define kmkernel KMKernel::self()
because 'kernel' was a much to general term. We really shouldn't repeat the mistakes of the X developers.
I noticed this problem when I played around with KImageEffects. kimageeffects.h contains 'kernel' as parameter of some methods and so the compilation had to fail. We won't need KImageEffects in the near future, but at least we are now prepared and a clash with another 'kernel' can't happen anymore.
svn path=/trunk/kdepim/; revision=252621
the job wasn't properly terminated, so the syncing 'state machine' was kind of stuck.
The underlying design issue is that KIO's connected slaves emit (via Scheduler) a
signal slaveError, totally unrelated to any job, and that killing the slave
kills the jobs _silently_ (without emitting result()).
As discussed with Waldo, the only way to handle this properly is to wait for the
slave to be connected (slaveConnected signal from Scheduler), before assigning
the slave to a job. I didn't rewrite all of kmail for it :) but I changed the
return value of makeConnection to be an enum: Error, Connected or Connecting.
This keeps compat with existing code which just checks it as a bool, but allows
KMFolderCachedImap to wait for the slaveConnected signal before proceeding.
Visually, this does the connection attempt at 0% progress instead of at 30%,
so it looks more logical :)
svn path=/trunk/kdepim/; revision=233801
- Enable inode/directory handling so cached IMAP folders with only
subfolders work
- Fix lots of comments
- Change the big server sync switch to carry on automatically instead
of always calling itself
svn path=/trunk/kdepim/; revision=217248
for me the "crash on reply" and "forward message crash" bugs and people
on IRC have been insisting on committing it. So yeah, asynchronous message
handling in KMail is in!!! May the God of Coding have mercy on us...
svn path=/trunk/kdenetwork/kmail/; revision=193386
Ingo Klcker
Bo Thorsen
Till Adam
Carsten Burghardt
Marc Mutz
Ladislav Strojil
Stephan Binner
Dirk Mueller
David Faure
Zack Rusin
Stephan Kulow
Antonio Larrosa Jimenez
Daniel Molkentin
Steffen Hansen