leap_mail.git - [leap

Age	Commit message (Collapse)	Author
2014-02-20	Fix regression on "duplicate drafts" issue.	Kali Kaneko
	Not a permanent solution, but it looks for fdoc matching a given msgid to avoid duplication of drafts in thunderbird folders.
2014-02-20	mail parsing performance improvements	Kali Kaneko
	Although the do_parse function is deferred to threads, we were actually waiting till its return to fire the callback of the deferred, and hence the "append ok" was being delayed. During massive appends, this was a tight loop contributing as much as 35 msec, of a total of 100 msec average. Several ineficiencies are addressed here: * use pycryptopp hash functions. * avoiding function calling overhead. * avoid duplicate call to message.as_string * make use of the string size caching capabilities. * avoiding the mail Parser initialization/method call completely, in favor of the module helper to get the object from string. Overall, these changes cut parsing to 50% of the initial timing by my measurements with line_profiler, YMMV.
2014-02-20	ignore keyerror on deletion	Kali Kaneko

2014-02-20	catch empty rdoc	Kali Kaneko

2014-02-20	catch stopiteration	Kali Kaneko

2014-02-20	fix rdoc duplication	Kali Kaneko

2014-02-20	fix attribute error on debug line	Kali Kaneko

2014-02-18	catch soledad error while updating mbox doc	Kali Kaneko

2014-02-17	avoid unneeded db index updates and rdoc creation	Kali Kaneko

2014-02-17	defer fetch-all-flags too	Kali Kaneko

2014-02-17	remove size calculation until we defer it to thread properly	Kali Kaneko

2014-02-17	cache uidvalidity	Kali Kaneko

2014-02-17	pep8 fixes.	Ivan Alejandro

2014-02-17	Update keymanager kwargs, related to #5120.	Ivan Alejandro

2014-02-17	Remove notify_new callbacks from fetch and copy.	Kali Kaneko
	This fixes a bug with qtreactor that was making the 'OK foo copied' not being delivered. This or something similar will probably have to be re-added, because on the current state the destination folder will not receive the notification if it's selected before the copy operation has finished. But in this way we have a clean slate that is working properly. The bottleneck in the copy/append operations seems to have moved to the select operation now.
2014-02-17	remove floody log	Kali Kaneko

2014-02-17	freeze dirty/new sets to avoid changes during iteration	Kali Kaneko

2014-02-17	defer message push to thread	Kali Kaneko

2014-02-17	speedup mailbox select	Kali Kaneko

2014-02-17	profile select	Kali Kaneko

2014-02-17	add cProfiler instrumentation	Kali Kaneko

2014-02-17	docstring fixes	Kali Kaneko

2014-02-17	avoid hitting db on every select	Kali Kaneko

2014-02-17	remove early notification on append for now	Kali Kaneko
	this can be done to save some msec, but additional measures have to be taken to avoid inconsistencies with reads right after this is done. we could make those wait until a second deferred is done, for example.
2014-02-17	suggest bigger threadpool to reactors that honor it	Kali Kaneko

2014-02-17	docstring fixes	Kali Kaneko

2014-02-17	select instead of examine	Kali Kaneko

2014-02-17	remove all refs during removal, and protect from empty docs	Kali Kaneko

2014-02-17	move mbox-doc handling to soledadstore, and lock it	Kali Kaneko

2014-02-17	purge empty fdocs on select	Kali Kaneko

2014-02-17	avoid revision conflict during deletion	Kali Kaneko

2014-02-17	defer appends too	Kali Kaneko
	and cut some more time by firing the callback as soon as we've got an UID.
2014-02-17	defend against malformed fdocs during unset dirty/new	Kali Kaneko

2014-02-17	defer fetch to thread	Kali Kaneko
	also, dispatch query for all headers to its own method.
2014-02-17	improve flag-docs relative internal storage	Kali Kaneko

2014-02-17	fix several bugs in copy/store	Kali Kaneko

2014-02-17	fix last_uid write to avoid updates to lesser values	Kali Kaneko

2014-02-17	fine grained locks for puts	Kali Kaneko

2014-02-17	separate new and dirty queues	Kali Kaneko

2014-02-17	catch typeerror too in empty definition	Kali Kaneko

2014-02-17	make the condition optional	Kali Kaneko

2014-02-17	call notify in reactor	Kali Kaneko

2014-02-17	fix repeated recent flag	Kali Kaneko

2014-02-17	do not get last_uid from the set of soledad messages	Kali Kaneko
	but always from the counter instead. once assigned, the uid must never be reused, unless the uidvalidity mailbox value changes. doing otherwise will cause messages not to be shown until next session. Also, renamed get_mbox method for clarity.
2014-02-17	add profile-command utility	Kali Kaneko

2014-02-17	defer_to_thread the bulk of write operations	Kali Kaneko
	and batch the notifications back to the memorystore, within the reactor thread.
2014-02-17	two versions of accumulator util	Kali Kaneko

2014-02-17	separate better dirty/new flags; add cdocs	Kali Kaneko

2014-02-17	remove hdoc copy since it's in its own structure now	Kali Kaneko

2014-02-17	make fdoc, hdoc, chash 'public' properties	Kali Kaneko