torshepherd wrote:

@HighCommander4 is right here. I have no emotional dibs on completing this 😅 
and it should be pretty easy to Claude up. But the current implementation had a 
major issue

- conflicting edits from apply all would cause the document to get modified 
into a broken state

And in attempting to fix that, I made it worse, where even nonconflicting edits 
now can incorrectly patch. How this should be resolved in the most rigorous way 
is still an open question: should we iteratively apply? What happens if two 
quick fixes are contradicting? What happens if one fix leads to more auto 
fixable things? Or should we just pick one of the fixes to apply in these cases?

If you wanted to pick this up, the first step would be to flesh out how we want 
all of these edge cases to work and write tons of test cases. The 
implementation after that should be trivial for a coding agent for instance.

But it comes down to reviewer burden ultimately

https://github.com/llvm/llvm-project/pull/79867
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to