copilot-cli hangs when terminal command larger then PTY buffer (4KB on MacOS)

[sysid]

Describe the bug

copilot-cli hangs when terminal command larger then PTY buffer (4KB on MacOS). It looks like a deadlock maybe because the copilot is too slow to drain/consume the PTY.

Affected version

No response

Steps to reproduce the behavior

Prompt: run 'seq 1 5000'

● Read shell output Waiting up to 10 seconds for command output
│
│ 1
│ 2
│ 3
│ 4
│ 5
│ 6
│ 7
│ 8
│ 9
│ 10
│ 11
│ 12
│ 13
│ 14
│ 15
│ 16
│ 17
│ 18
│ 19
│ 20
│ 21
│ 22
│ 23
│ 24
│ 25
│ 26
│ 27
│ 28
│ 29
│ 30
│ 31
│ 32
│ 33
│ 34
│ 35
│ 36
│ 3
│ <command with shellId: 0 is still running after 10 seconds. The command is still running. Check the output above - if it looks like the command is waiting for input, use write_bash to provide it. Otherwise, use read_bash to continue waiting for output, or stop_bash to stop it.>
│
└ 38 lines

◐ This is strange. seq 1 5000 should complete in milliseconds. It seems like it's stuck somehow. Let me stop it and try a different approach.

Expected behavior

command should return within milliseconds and output be given by copilot. No blocking.

Additional context

No response

[kexinjinnn]

Seeing this too — the CLI hangs when a tool command produces a lot of output, and never returns control; the turn sits on "Thinking" and Escape then hangs on "Cancelling," so I have to kill the terminal (copilot --resume recovers).

The PTY-buffer-drain deadlock described here looks like it could be the root cause behind a cluster of intermittent hang reports: #2703, #2770, and the earlier #405 / #529 / #575. #529 was traced to abort signals not being propagated into the streaming loop, which feels adjacent to a slow-drain problem. I can't give a 100% reliable repro since it's timing-dependent, but this seems like a high-impact one. Happy to attach a --log-level debug session log next time it reproduces.

[szabta89]

Fixed starting 1.0.60. Please create a new issue if you observe further problems.