the test failure is not the fault of this diff. the coverage job did pass, the failing is because of lilybuild.

Knowing whether an outgoing event is sent is of no use. It seems that we should add the outgoing events right away.

Only keep codes into model, leave pending events as return value

Use failWithResponse(r) instead of lambda

Make breakpoints in Client::startSyncing() const

Check if it still crashes if we cut off network after opening

The RoomModel in libkazv should remember more than the current text-only localDraft. Should also store the state of e.g. editing or replying.

Pinned messages page: Modify the behaviour so that when the pinned messages action is triggered, the pinned messages page will replace the room page. Then, add an action in pinned messages page to return to the room page. Leave it that way until we refactor PageRow in T182 .

Some pages are left unmodified.

Move display codes and pending events into VerificationProcess

In D245#5654, @nannanko wrote:

Currently Kazv cannot remember a room is in the state of editing or replying to an event, which makes restoring a draft difficult to achieve. I think this feture should be divided into several diffs instead of being completed in one diff.

Currently Kazv cannot remember a room is in the state of editing or replying to an event, which makes restoring a draft difficult to achieve. I think this feture should be divided into several diffs instead of being completed in one diff.

Show failure message

better tests and error handling

What if one party sends a bad key event? Should test that it correctly sends a cancel event of invalid_message

I don't think we want to do this. The benefits of keeping Crypto a value type isn't great. For example, it delays saving of the crypto state, which can cause problems in the case of a crash / kill of the process. A solution might be to make Crypto a dependency of the store instead.

mmore tests and error handling