[BUG] High incidence of grey screen of death (GSOD) from some update this week

[nabilfreeman]

Problem (one or two sentences)

Hi, just reporting that I'm getting multiple grey screens of death (GSOD) a day for the last couple of days. Something is locking up the CPU in the VSCode window when the GSOD happens because it becomes completely unresponsive.

Context (who is affected and when)

I'm on the latest version of Roo and this issue has been happening for at least the last 2 updates.

Other Roo instances in other VSCode windows continue to work, but the VSCode diff editor reports timeouts across all windows which breaks execution in other Roos as well.

After about 2 minutes of GSOD, the VSCode window crashes with error "The window terminated unexpectedly (reason: 'crashed', code: '5')"

After pressing the "Reload" button, the affected window works again, and diff editor is functional again in other Roo instances too.

I haven't tried a rollback yet but if it continues I'll try and find and older version that eliminates the issue to narrow things down further. For example, downgrading to 3.28.latest instead of 3.29.x

Reproduction steps

1. Use the latest version of Roo Code
2. Open multiple worktrees of the same project in different windows (2 active windows is enough to replicate)
3. Run Roo Orchestrator workflow for two different tasks, one on each worktree
4. After some time of normal functionality, one of the Roo Instances will GSOD, the window will freeze completely. The other Roo instance will continue to work, but its file edits will fail due to an error saying the diff editor is unavailable

Expected result

Roo is usable in multiple VSCode windows at the same time

Actual result

Roo is mostly usable but randomly crashes

Variations tried (optional)

No response

App Version

3.29.4

API Provider (optional)

None

Model Used (optional)

No response

Roo Code Task Links (optional)

No response

Relevant logs or errors (optional)

Here is the diff editor error I get in the still-alive Roo instance:

Error writing file:
Failed to open diff editor for /redacted/file.ts within 10 seconds. The editor may be blocked or VS Code may be unresponsive.

[nabilfreeman]

Update, I just got it again since opening this issue. It happens as soon as a terminal command is executed by a sub-task. I will stay on this version for the rest of the day to try and collect more information but then I will have to downgrade as the interruptions are very cumbersome.

[nabilfreeman]

My Roo Terminal settings:

[roomote]

I'm exploring a potential fix for this GSOD issue. The problem appears to be related to terminal command execution in sub-tasks causing VSCode windows to become unresponsive when multiple Roo instances are running. I'll investigate the terminal handling and diff editor synchronization code to identify the root cause.

[roomote]

I've opened PR #8950 with a fix for the GSOD issues: #8950

The fix addresses the root causes by:

• Making terminal shell execution handlers non-blocking to prevent UI freezes
• Adding retry mechanisms for diff editor operations
• Implementing proper task isolation to prevent cross-window interference
• Including comprehensive tests for concurrent operations

Initial CI checks are passing and the full test suite is running. This should resolve the freezing and crashing issues when using multiple Roo instances across different VSCode windows.

Feedback and testing welcome!

[nabilfreeman]

I don't think that PR gets to the root cause, just my 2c. I also had a grey screen right after pressing enter to send a message which is nothing to do with Terminal.

Loads more crashes, so I have downgraded to 3.28.18. I'll update if the problem has stopped, which may help narrow it down to some diff in the 3.29.x commits.

[daniel-lxs]

Hey @nabilfreeman thank you for reporting this issue, we are looking into it!

[nabilfreeman]

Thanks @daniel-lxs . I'm on the older version now and things have been stable for the last 90 mins of work. Maybe it would be useful for Roo to have a logfile that I could dump when the gray screen happens. Obviously it would only be showing things up to the moment of the crash but it might be useful to pinpoint the issue as it seems to be a long term thing with multiple causes.