Sign in to confirm you’re not a bot
This helps protect our community. Learn more

AI After Hours | Debugging Unit Tests with GitHub Copilot

Microsoft Visual Studio
335K subscribers
<__slot-el>
<__slot-el>
1.5K views 11 months ago
In this episode, John Fleisher and Wendy Breiding from the testing team preview their product vision for using GitHub Copilot to help debug unit tests in Visual Studio. From conversational debugging to highlighting key object values at each breakpoint, there's lots to unpack here. GitHub Copilot is a coding assistant powered by Artificial Intelligence (AI), which can run in various environments and help you be more efficient in your daily coding tasks. == Introduction == 0:00 - Developers developers developers. 1:30 - Developers always debug tests, we want to help them with that.  == Demo == 2:10 - Debug failed unit test with GitHub Copilot. 2:35 - GitHub Copilot produces a debugging plan. 3:00 - Important object values shown in text editor at breakpoint. 3:19 - GitHub Copilot is tracking object values. 3:27 - GitHub Copilot suggests a fix. == Discussion == 4:00 - John Fleisher has worked on test infra for 5-6 years. 4:30 - Developers see …...more
...more

Developers developers developers.

0:00

Developers always debug tests, we want to help them with that.

1:30

Debug failed unit test with GitHub Copilot.

2:10

GitHub Copilot produces a debugging plan.

2:35

Important object values shown in text editor at breakpoint.

3:00

GitHub Copilot is tracking object values.

3:19

GitHub Copilot suggests a fix.

3:27

John Fleisher has worked on test infra for 5-6 years.

4:00

Developers seem to want help with diagnosis more than they do with test generation.

4:30

Context package includes stack frame, symbol information, test failure information.

5:10

With a Git repo attached, GitHub Copilot also considers code changes to test and related symbols when diagnosing test failure.

8:03

Developers are encouraged to ask followup questions to chat when debugging failed unit test.

10:00

We think this is particularly useful when developers are unfamiliar with the codebase/test.

11:33

More targeted guidance for specific exception types could improve debugging guidance in the future.

14:44

Developer is always in control of the debugger.

16:00

Test team is curious about further integrations with debugger, tool calling, and database content.

17:55

More context is not always better.

24:25

Please leave feedback about scenarios from current "Ask Copilot" for a failed unit test that do and don't work well via "Report a Problem" in Visual Studio.

26:00
AI After Hours | Debugging Unit Tests with GitHub Copilot
20Likes
1,517Views
May 222024
In this episode, John Fleisher and Wendy Breiding from the testing team preview their product vision for using GitHub Copilot to help debug unit tests in Visual Studio. From conversational debugging to highlighting key object values at each breakpoint, there's lots to unpack here. GitHub Copilot is a coding assistant powered by Artificial Intelligence (AI), which can run in various environments and help you be more efficient in your daily coding tasks. == Introduction == 0:00 - Developers developers developers. 1:30 - Developers always debug tests, we want to help them with that.  == Demo == 2:10 - Debug failed unit test with GitHub Copilot. 2:35 - GitHub Copilot produces a debugging plan. 3:00 - Important object values shown in text editor at breakpoint. 3:19 - GitHub Copilot is tracking object values. 3:27 - GitHub Copilot suggests a fix. == Discussion == 4:00 - John Fleisher has worked on test infra for 5-6 years. 4:30 - Developers seem to want help with diagnosis more than they do with test generation. 5:10 - Context package includes stack frame, symbol information, test failure information. 8:03 - With a Git repo attached, GitHub Copilot also considers code changes to test and related symbols when diagnosing test failure. 10:00 - Developers are encouraged to ask followup questions to chat when debugging failed unit test. 11:33 - We think this is particularly useful when developers are unfamiliar with the codebase/test. 14:44 - More targeted guidance for specific exception types could improve debugging guidance in the future. 16:00 - Developer is always in control of the debugger. 17:55 - Test team is curious about further integrations with debugger, tool calling, and database content. 24:25 - More context is not always better. 26:00 - Please leave feedback about scenarios from current "Ask Copilot" for a failed unit test that do and don't work well via "Report a Problem" in Visual Studio.

Follow along using the transcript.

Microsoft Visual Studio

335K subscribers