KDE Frameworks 5.70 was just released and should be trickling out to users of rolling release distros at any time. Various Arch users who have already received the update have been complaining about slow animations in Plasma, and I wanted to write a blog post to explain what’s going on here. It is a bit technical so let me start with the TL;DR version: “releasing software is complicated and this will be fixed once Plasma 5.19 comes out next month.”
For the longer version, allow me to explain:
This is caused by an unfortunate timing problem stemming from the different Plasma and Frameworks release schedules.
Plasma and Kirigami-based apps use standard duration values for animations (e.g. units.shortDuration, units.longDuration, etc.) to keep animation timings relatively consistent. These duration values are set in the respective Frameworks: plasma-framework and kirigami.
I recently discovered that Plasma units were far shorter than Kirigami units. For example a Kirigami units.longDuration unit is 250ms, while a Plasma units.longDuration unit was 120ms–over two times faster. A Kirigami units.shortDuration unit was 150ms, while a Plasma units.shortDuration unit was 24ms–almost too fast to see. In practice the Plasma units.shortDuration value was useless and always had to be multiplied by something. Even most of the longDuration values were being multiplied by random numbers too. So we wound with animated transitions throughout Plasma having timings like units.shortDuration * 4 or units.longDuration * 3. It was a classic problem of badly-chosen library constants that force apps to work around them and munge them this way and that, totally defeating defeated the purpose of using standardized values in the library in the first place. There was not actually any standardization at all!
I needed to fix this as a part of my introduction of a new animation-using component, the ExpandableListItem (which I keep meaning to blog about): https://phabricator.kde.org/D28033
I fixed the Plasma units to be the same as the Kirigami units in https://cgit.kde.org/plasma-framework.git/commit/?id=0739113a4477e1eb25bf13b0040af5a502d3ef0a, and then fixed Plasma itself to no longer multiply the units in a series of other commits. However this presented an issue: Plasma and Frameworks have different release schedules! So people will not get both aspects of the change at the same time! This means that for a time, some people would have animations that were undesirably slower or faster. How should this be handled?
Unfortunately there is no easy way to do conditionals depending on a frameworks version in QML code as we can in C++ code, so that easy option was not available. Probably something to look into implementing.
So we had a few options. One was to avoid solving the problem until Plasma 6, several years in the future, at which point we could do everything at once. This was not deemed satisfactory, as the issue was blocking the ExpandableListItem patch which was needed for a task targeted for Plasma 5.19. Another option was to leave the existing units alone for Plasma 5, and add new units with different names now, and have Plasma 5.19 use those new differently-named units. This would have avoided the issues you’re all experiencing, but would have resulted in terribly confusing code. In the end we decided to spare ourselves the potential for new bugs stemming from that.
The final option was to wait to make the Frameworks change in a Frameworks release that lines up as closely as possible with the Plasma 5.19 release. Plasma 5.19 depends on Frameworks 5.70, but always releases about a month later, at which point Frameworks 5.71 will be out. This option therefore presented two sub-options: put the units change in Frameworks 5.70 or 5.71?
If we did it in 5.70, there would be a one-month period in which people using rolling release distros suffer from slow animations, because they have Frameworks 5.70 but not Plasma 5.19 yet.
If we did it in 5.71, the period of time in which people suffered from this issue would still exist, but it would potentially be shorter. However depending on distro release schedules, if a distro released Plasma 5.19 *before* Frameworks 5.71, then animations would become too fast to see! Furthermore, any discrete release distro in the future that shipped Plasma 5.19 with the 5.70 Frameworks version it depends on rather than a newer one would then have all of its users suffer from the bug forever (or at least until its packagers backported the plasma-framework commit).
So shipping the units change in Frameworks 5.71 did not seem to be a realistic option. In the end I shipped the units change in Frameworks 5.70 knowing that rolling release distro users (myself included) would suffer from slow animations for one month. Sorry. 😦 It will all be fixed in Plasma 5.19.
Software is complicated!