-
Notifications
You must be signed in to change notification settings - Fork 2.8k
Description
Problem (one or two sentences)
When Roo is streaming long thinking blocks (and possibly long mode handoff messages) scroll-to-bottom can get disengaged. The long block may even scroll to bottom for a while and then disengage scroll-to-bottom before it's completed.
Context (who is affected and when)
I'm experiencing this issue when running multiple instances of VSCode and the system as a whole is under high CPU load. It's not clear whether the system load is significant, but the issue is frequent enough to be very visible when monitoring multiple VSCode instance for Roo Code extensions needing intervention or attention.
Reproduction steps
No clear repro of the issue yet other than running many sessions for extended periods of time and needing to frequently scroll to bottom to see if an instance needs attention despite ensuring all instances are scrolling to bottom at various points in time.
Expected result
Thinking blocks that started when scroll-to-bottom is engaged continue to scroll to bottom throughout.
Actual result
Thinking blocks start will start when scroll-to-bottom is engaged, scroll-to-bottom as expected for a while, but a high percentage of the time will stop scrolling to the bottom before long thinking blocks are complete.
Variations tried (optional)
No response
App Version
3.38.2
API Provider (optional)
Not Applicable / Other
Model Used (optional)
Roo Code Cloud/GPT-5.2 to see lengthy thinking blocks
Roo Code Task Links (optional)
No response
Relevant logs or errors (optional)
Metadata
Metadata
Assignees
Labels
Type
Projects
Status