Previously the print job was sent with multiple PUT requests. This was necessary since the request was made with the Python Requests library, which freezes the interface while it's running. By breaking the upload up in chunks, it would at least periodically keep updating the interface (4 times per second or so, depending on your internet speed and Ultimaker's). Later on, this was replaced by a QNetworkRequest which doesn't freeze the interface in a refactor, but it was still broken up in chunks.
Recently that was again refactored to use Uranium's HttpRequestManager, which under the covers also uses QNetworkRequest. This also doesn't freeze while uploading, which is good. However for some reason the second chunk would always give a QNetworkReply::ProtocolUnknownError, which is a bit of a catch-all error for any HTTP status codes not supported by Qt. I was not able to figure out what was really going wrong there, but it has something to do with the content-range header and making multiple requests.
Instead of fixing that, I've removed the chunking. This simplifies the code a lot and doesn't have any implications for the user, since it still doesn't freeze the interface while the network request is ongoing. It should be slightly faster to upload, and reduce load on the server too.
However the disadvantage is that the original bug is probably still there if it was a bug in the HttpRequestManager. If it was a bug in the ToolPathUploader I probably deleted the bug here.
Contributes to issue CURA-7488.
Surround the serve_forever function of the web server with a try-catch
on Windows, in order to avoid printing the entire (useless) traceback.
Now a warning message is be displayed in the log instead.
The behavior is untouched in other platforms
I have no idea why i thought a listview was needed for this when
this was made. The data shouldn't be flicable, so it makes way more sense to
use a row layout
CURA-7480
These two version upgraders do the same operation. Their code is copied. Instead, we can just let one version upgrader upgrade both versions of the files.
Contributes to issue CURA-7413.
This is necessary because our code now converts this to an integer. Mocks can't be converted to an integer, so the sub-magic-mock here can't be used. We need to provide something that can be cast to an integer.
Otherwise the list of extruders becomes a MagicMock. Adding a list to a MagicMock is again a MagicMock, and when iterating over that it doesn't result in any iterations. As a result, the for loop that clears all user containers didn't clear anything and the test failed.
If Z hop is enabled, it would move to the height of the Z hop at the start of the print, in order to hop for the initial travel move. Cura sort of assumes that this hop is going up and so it's safe to do, but in reality this hop is going down due to the starting process in the firmware. As a hotfix, we'll just move away from that blob so that we don't hit it even if you enable Z hops.
