I filed this issue about a regression in semantic highlighting about four months ago. I worked with someone on it about a month ago who asserted that it was due to an inspection in my plugin, but as I’ve stated in the issue multiple times, it’s consistently reproducible with all inspections disabled.
I continue to have it reported by users of my plugin, and I don’t really have anything useful to tell them since it seems to be a regression starting in 2025.3+ with the exact same build of my plugin installed.
It’s certainly still possible that it’s an issue in my plugin, but given that it works fine in 2025.2 and earlier (and has for over a decade), and given that it’s consistently reproducible with all inspections disabled, I’m unsure how to proceed.
I’ve asked for a status update a few times over the past month, but it seems that the JetBrains side has gone dark on it.
I understand that it’s perhaps not as critical to JetBrains as many other items, but I’ve just not seen activity go dark on a support issue like this in the past.
I guess the best way is to ask in the issue where the team and support engineers read comments
I agree 100%, and I’ve done exactly that twice over the past month with no response. Should I just keep asking for status in the context of that issue on a weekly basis? At what point does the lack of additional feedback motivate some other channel of communication?
We only have so many engineers on the team and as I see the real engineer who actively develop the subsystem answers where when they are available. As far as I understand, we could not reliably reproduce in our side, it seems there is no test / reproducer for us to debug
I understand lack of resources. My understanding of the last known status from the JetBrains side is that the engineer said it was attributable to my plugin’s unused declaration inspection, but I’m able to reproduce it 100% consistently with that inspection (and all inspections) completely disabled. Since raising that point, I have not received any further update as to whether or not that engineer sees the same behavior or not. It’s just been complete silence. It seems like that does warrant at least some status update, no?
A viable way forward would be extracting a minimal reproducer code out of your plugin, so it could be debugged and observed
I’ve considered that, but “minimal” is going to be pretty extensive in practice since it requires a custom language with annotator-based semantic highlighting being displayed in an EditorTextField component that’s not backed by a physical VirtualFile.
It is consistently reproducible with my plugin installed in a quite simple and straightforward manner as described in detail in the linked issue.
I see, but we have no any chance to debug it and extract a test
Hmmmm…interestingly it looks like this issue may have been fixed in 2026.2.0.1 or 2026.2.1. It seems to be behaving properly again in those versions but not in 2026.1.4 or 2025.3.6.1. I know it was still present in 2026.1 GA.
Investigating further to see if I can confirm that it’s actually fixed and, if so, nail down an exact update that fixed it so I can look at what might have changed in that area.