Is there an automated way to always try to fix the code after failed tests instead of What's wrong? Fix queries?

[Dima-369]

Issue

I have aider break my tests often, and it would be so amazing if it could just follow up itself to failed tests to fix the code, instead of prompting me with What's wrong? Fix queries. Maybe with a configurable limit of, say 3, failed tests in succession to finally ask the user for help.

Maybe with a new setting auto-test-fix: true (or auto-test-fix-limit: 3) for auto-test: true?

What do you think? Cline behaves this way, and it often manages to fix the tests itself, which is great.

Version and model info

Aider v0.78.0
Model: openrouter/deepseek/deepseek-chat:free with diff edit format, prompt cache, infinite output
Git repo: .git with 272 files
Repo-map: using 4096 tokens, auto refresh
Added memory-bank/techContext.md to the chat.

[Dima-369]

I think that might be a bug in the code. cmd_run() never returns anything, so cmd_test() also returns nothing, so test_errors = self.commands.cmd_test(self.test_cmd) is never filled, and the tests are never tried to be fixed automatically.

cmd_run() should return msg on errors, then it works.