If KRunner has multiple threads accessing the same sqlite database, then yes.

Vault uses FUSE, no block-level for the time being. I was toying with Tomb and preallocated fixed-size containers in the filsystem which uses Luks in the back, but I'm not sure it can be made to work properly.

Vault does not support public-private key encryption - it is out of the project's scope. And it would be quite strange for the user to see items like "KMail Index for ABCDEFGH01234... key" in the list of existing vaults. :)

"can't" should not be used in formal writings. Not sure how formal documentation is considered to be.

You can also switch to 'can't' as far as I'm concerned

The fix needs to be a bit more complex. We can not really rely on the user to fix the database.

Well, in a library that 99% relies on a database, having no database is not something that can be gracefully handled (the returned nullptr is asserted on as well in the function that calls ::instance in ResultSet :) - because of this, I'm changing to 'Request Changes').

This is not really as simple as 'accept without a comment'.

I can not say with certainty - never tested html encoding, but I would expect it to work in general.

Maybe hardcoding is a good idea, I have nothing against that. Though, I'm not sure hardcoding to 6 is the best one - I'm torn between that one and the next :)

Noise is often used to break the gradient visual artifacts that appear due to screens' inability to show colors precisely. You often get the impression there are lines orthogonal to the gradient direction which ruins the gradient.

Otherwise +1

I've pushed this patch with a few changes (don't want to wait the last moment before the string freeze).

Hi @abetts , we need a nice dialogue that shows the password entry for unlocking a vault. You have the current proposal screenshot above.

Since VDG has not responded. Can you remove the padding and make the name bold?

Why don't you set up the preferred size for the dialogue, and allow the label to be multi-line?

@ngraham I can not reproduce the text overflow bud you are getting. I'm not setting any size-related things, so the bug is probably in the upstream (libplasma or above)

• Reverted the 'create a new vault button' to be a proper button

Almost mergeable - I just have a few nit-picks:

I was on a vacation, sorry. I'll test everything today and merge it if I don't find a problem.

The activities cache is a private class meant to communicate directly to KAMD. The signals it contains are in the form (activity id, something) and are meant to be private. The KActivities::Info class was meant to wrap these signals into a nice OO API where an activity object (Info) sends signals that a particular property has changed.

While I like the idea behind the patch, I'd rather have what David proposed. Namely, d_ptr is a smart pointer, but I don't think it should be this smart - creating specializations for different types' destructors should not really be a part of it.

Also, not sure adding a signal that has one user only to a shared library is the best approach, especially since .h says this:

This one is almost mergeable.

I was not clear enough probably - I meant that "Have a default backend" does not warrant a change to the UI. Especially when it makes the UI more complex.

p.s.

Do I need to relogin to make a kded module show up?

Having a default backend (IMO) has nothing to do with the UI. What is the rationale behind changing a combo-box (a widget designed for multiple choice questions) to a button?

Looks ok.

Somehow I didn't submit the comment when I wrote it. See above.

Increasing the result limit

More debugging...

Removed model test

p.s. This is how I restart everything:

Another one to test:

• fixed a small bug related to fav adding
• added debugging output for easier testing (when you encounter a bug, send me an email with plasma output)

Fixing the ordering

Post-migration sorting still feels completely random - not sure how solid that feeling is though.

The last patch supports Kickoff, Kicker and Dash. I've tested what I can, found an small issue in KAMD (update KAMD to latest master before testing) and it looks like we are on the finish line.

Good news: There's actually nothing to do on the favorite ids, Kicker already uses
KService::storageId as favorite id, which will be the menuId when it's available.

This might be a hint:

Regarding Kickoff - it looks like there are some problems in its model handling:

The initial sorting of the migrated superset seems random. How is it done?

With icons (designed after the Dolphin plugin for activity linking - to at least be consistent with something :) )