This adds a check before the storeRoomUpdate() call
if the room is actually known. This has the effect
that the call of forgetRoom() is skipped.
The reason for this is an edge case in the database
implementation when calling getAllKeys(). This
somehow can corrupt the keys-cache and lead to
some problems. I wasn't able to fix this problem
yet so this refactoring is more a good-enough
workaround for now to not trigger it on an
initial sync. I plan to fix it with a different
approach which completely removes the keys-cache
in the future.
However this change leads to some problems in the
tests as they already rely on this edge case.
- unify behavior of all message sending related command
- add a StringBuffer as stdout-like output buffer for commands
- create a typedef for the command function signature
- create a common exception type for command execution
- enable commands to run on Client-level rather than Room-level
- BREAKING: Client.addCommand signature now takes an optional StringBuffer as second parameter
We actually never use the
roomId of this type and the
matrix spec has changed in
a way that the roomId is
never sent there actually. So
it was super easy to just
replace all BasicRoomEvent
with BasicEvent. The roomId
became nullable anyway.
This also makes sure that
room account data does not
get unnecessarily serialized
and deserialized before
storing it in the database.
For this it changes the
code flow at multiple
places.
This changes the database
api a little bit so that it does
not unnecessarily serialize
and deserialize all
account data objects. Should
improve the performance of
the SDK.
This handles room ephemerals
directly and not as room
event updates, which saves
one unnecessary step to json serialize and
deserialize. Handling those
as room event updates had
no benefit anyway so this
should speed up performance.
This fixes a huge performance
leak as for every encrypted room
the whole member list is loaded
every time you start the client.
This also changes the default value of `cache` in
this method to if the room is encrypted or not.
For encrypted rooms we always want a complete
member list locally.
Covers a few edge cases that still fail. Changes to the unlocalizedBody
function shouldn't cause behavioural changes apart from fixing a few
edge cases.
It is not clear why we ever would want to return the formatted_body when
we ask for the body, but it seems to not be used anywhere and there are
no tests covering that functionality. However it leads to suprising
results, where the plaintextBody can be tricked into returning html
without applying conversions. So we just get rid of that functionality.
Christian Kußowski
krille-chan
Karthikeyan S
td
Yash-Garg
Mohammad Reza Moradi
The one with the braid
Nicolas Werner