| Age | Commit message (Collapse) | Author |
|
Setting PRAGMA synchronous = 0 is really unsafe, and leads to database
corruption (which I've personally experienced). This commit uses
SQLite's Write-Ahead Log (WAL) [1] instead, together with synchronous =
NORMAL. According to [1], this trades off performance for durability
(i.e. it's possible that some transactions may not have committed if the
power gets lost), but still guarantees that the database won't corrupt
itself.
Together, these changes should improve reliability whilst either
improving or having no effect on performance.
[1]: https://www.sqlite.org/wal.html
|
|
|
|
|
|
Fixes #911
|
|
|
|
|
|
|
|
|
|
fixes #831
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
list (#764)
Co-authored-by: codedust <codedust@users.noreply.github.com>
Co-authored-by: fiaxh <git@lightrise.org>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
fixes #495
|
|
|
|
resource we asked for)
Some servers give out random resources. Then, we didn't deduplicate own messages afterwards because we saved the wrong resource.
|
|
iterating over it)
|
|
|
|
|
|
|
|
|
|
request_info/items async, add caching has_entity_feature
|
|
|
|
fixes #756
|
|
fixes #179
fixes #812
|
|
|
|
Fixes #791 where a signal to update the UI is sent to
late when sending files, resulting in an out of sync
conversations list.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
fixes #591
|
|
|
|
|