Bug 340698

Summary: Composer: Inconsistent UI indication when setting a reply mode when the original message disappeared
Product: [Unmaintained] trojita Reporter: Jan Kundrát <jkt>
Component: Desktop GUIAssignee: Trojita default assignee <trojita-bugs>
Status: RESOLVED FIXED    
Severity: normal    
Priority: NOR    
Version: git   
Target Milestone: ---   
Platform: Other   
OS: Linux   
Latest Commit: http://commits.kde.org/trojita/46f79b80fbe1aae3f1d3f45851644bc07630dbf2 Version Fixed In:
Sentry Crash Report:

Description Jan Kundrát 2014-11-06 23:02:52 UTC 
- start replying to a message,
- change the replying mode to "hand-picked recipients"
- go offline
- go back online (this invalidates the QPersistentMessageIndex)
- change the replying mode to something else
- note that the combobox still shows "hand-picked recipients" (which is correct as the index is no longer valid), but upon expanding the menu, the entry which you selected is shown as checked
Comment 1 Thomas Lübking 2014-12-14 11:44:15 UTC 
Is there intent to ever support reply modes across offline sessions?

Otherwise imo the proper fix was to catch the connection loss (or general invalidation of m_replyingToMessage) and switch to handpicked + disabling the button.
Comment 2 Jan Kundrát 2014-12-14 11:54:33 UTC 
I think it "would be great", but there's a substantial amount of code which needs deep changes to make it happen.

As you said on Gerrit -- intercepting a failure of the setReplyMode followed by a check for index validity and if it's gone, switching unconditionally to the handpicked mode and disabling others should do the trick.
Comment 3 Thomas Lübking 2014-12-14 14:46:00 UTC 
handling the item is trivial - i link to the models dataChanged signal and when the index is invalid afterwards, fix the GUI.

About preserving the message:
is it too naive to simply store model, mailbox name and UID and when the model changes (for the better ;-) try to lookup the index as ImapMessageAttachmentItem does?
Comment 4 Jan Kundrát 2014-12-14 14:53:04 UTC 
(In reply to Thomas Lübking from comment #3)
> is it too naive to simply store model, mailbox name and UID and when the
> model changes (for the better ;-) try to lookup the index as
> ImapMessageAttachmentItem does?

Right now, the whole tree gets invalidated during reconnect, and this is needed due to the way we handle TreeItem::UNAVAILABLE, LOADING and any other errors. Mailbox discovery happens asynchronously, and each level of mailbox nesting gets discovered on demand.

I'm not saying this isn't doable, but I do say that it's hard.
Comment 5 Jan Kundrát 2014-12-19 21:14:17 UTC 
Git commit 46f79b80fbe1aae3f1d3f45851644bc07630dbf2 by Jan Kundrát, on behalf of Thor Nuno Helge Gomes Hultberg.
Committed on 16/12/2014 at 21:30.
Pushed by gerrit into branch 'master'.

Fix inconsistent checkbox state

Solves contradiction regarding information about replying mode when the
message that is being replied to disappears.

Change-Id: I66b265d33711107fd16cfd13192f848228f6adcb

M  +19   -9    src/Gui/ComposeWidget.cpp

http://commits.kde.org/trojita/46f79b80fbe1aae3f1d3f45851644bc07630dbf2