[BUG]: eval "$(gh copilot alias -- bash)" takes 0.5 second to load

[yermulnik]

What happened?

What did you do? What happened? What did you expect to happen?

eval "$(gh copilot alias -- bash)" took 0.5 sec to load (I'm the "User is not authorized to use this Copilot feature"). I'd expect it to load much faster as 0.5 sec is a way too slow.

Versions

What versions of the relevant software are you running?
For example: gh v2.37.0, gh-copilot v0.5.2-beta

• gh version 2.52.0 (2024-06-24)
• version 1.0.3 (2024-05-08)

Relevant terminal output

Please copy and paste any relevant terminal output.
Please check your output before submission to ensure sensitive information is redacted.

> time eval "$(gh copilot alias -- bash)"

real    0m0,611s
user    0m0,210s
sys     0m0,053s

and another workstation:

> time eval "$(gh copilot alias -- bash)"

real    0m0,548s
user    0m0,156s
sys     0m0,037s

Ubuntu and Mint if it matters.

[jarrodldavis]

I have the same issue using eval "$(gh copilot alias -- zsh)" on macOS. Strangely, the entire script to be evaluated is emitted instantly, it's just that the gh command takes a long time to exit.

[BagToad]

I suspect this actually has more to do with gh startup time than the gh-copilot extension - see cli/cli#9724

@yermulnik @jarrodldavis, could you try setting TCELL_MINIMIZE=1 and let us know if that improves the performance?

[jarrodldavis]

@BagToad That helped performance a bit, but only reduced from 0.11s to 0.07s (according to time). As I noted initially, the issue is after the script is emitted, suggesting that it's an issue with the exit time, not the startup time. Additionally, I notice this slow exit consistently across all gh copilot commands (even gh copilot --help), but the root gh command is fast.

[yermulnik]

> export TCELL_MINIMIZE=1
> time eval "$(gh copilot alias -- bash)"

real    0m1,250s
user    0m0,167s
sys     0m0,080s

# second run
> time eval "$(gh copilot alias -- bash)"

real    0m1,116s
user    0m0,120s
sys     0m0,037s

🤔

[BagToad]

Ah, thanks for trying that out! Sounds like we have more investigating to do 😁