Looking at the code base, there are two steps that should be taken before any radical change occurs:

given that I've tested fastmail and we have not recieved further debug output (such as from "sinksh list folder") I'm assuming it's also a subscription issue.

at least on mailbox.org there is the problem that INBOX is not subscribed by default (which seems like a weird choice).
To counter that kube now ignores the subscription state for inbox and always enables it.
Building a new macos image is still pending.

Looks good, but please rename the folder to lowercase middleclickloader.

This is what approach 1 looks like with (clockwise) Breeze Dark, Oxygen and Wonton Soup:

Should this show an error to the user? I could also imagine this interfering with thumbnail generation when viewing the device

@apol I would like your opinion before merging

I plan to add a progress dialog and the ability to cancel.

Clicking again will undo the zooming action (or zoom to 100% if there is no previous zoom level) and unpress the button.

Let's land what we have, and continue working on the rest after that.

It was because of gpg's auto-key-retrieve feature. Not much we can do about it.

Add Logout function using krunner in basetest

• Restore TabBar visibility and position settings
• Connect to KonsoleSettings directly
• Remove Borders
• Add the Close and New Tab Buttons
• Re enable close tabs, simplify code
• Focus widget in currentChanged
• Re enable rename tab
• Remove replay signal.
• Fix Rename of tabs
• Re-Enable dragging tabs
• Re-Enable detaching actions

I guess this only happens if Telegram wasn't running already.

In D6083#114550, @gateau wrote:

I love the idea, but what about caching? Cached thumbnails are supposed to follow https://specifications.freedesktop.org/thumbnail-spec/thumbnail-spec-latest.html#THUMBSIZE, which still, sadly, says 256.

Done!

Map cursor position to current image
Limit cursor centering to image view

In D14093#292918, @rkflx wrote:

• In Compare mode it does not work right for every document. Adding mapFromScene fixes it for me.

I am fine with the content, but do fix the markup.

it looks correct to me (i. e. we should never delete a Folder without invalidating), but it would be nice with at least a shortened backtrace because I'm still not 100% sure how it ends up in that state.

please fix what harald pointed out, otherwise it looks good.

Looks like it works indeed :-)

Nice! Tested it:

• Working great for Fit.
• We should really add the fallback for !view->underMouse(), otherwise it's just too confusing.
• In Compare mode it does not work right for every document. Adding mapFromScene fixes it for me.
• For Fill, clicking on the bottom of a tall and completely visible image fits to width, but centers instead of panning to the bottom parts. (Not a blocker to prevent the patch from making it into the Beta on Thursday, but nonetheless maybe you can find the issue, which might be in the zooming code for filling.)

In regards to Python, that is not enabled as the bindings generation is not reliable and can easily fail on a highly parallel system (which the CI is, depending on the node it could be a Hexa Core with HT, so run using make -j13)
The generation also had a nasty tendency to fail for certain Frameworks more often than not due to various generation failures which would disappear on subsequent runs.

I have tested your patch: painting benchmark passes here, painting seems to work as well.

I would prefer those keyword changes to be done separately

This is on my radar and looked good from a diff review. I'd like to give this a proper review though and that will take a bit as I am super swamped just now.