Description
Same as #10360, reportedly fixed. Still/again happening sometimes. I'm not sure when.
Some of the reported problems by the LLM (GPT-5.1):
The patch failed because the exact header text has diverged.
The patch failed because my hunks used prefixed line numbers from the reader instead of matching exact file content
The patch failed due to line-number anchoring
The patch failed because my line-number-based hunk didn’t match the current file contents.
The patch failed due to an exact-line mismatch, likely because of quoting differences.
The patch failed due to slightly different import ordering/spacing than I assumed.
Maybe the model just produces illegal patches?
Plugins
None
OpenCode version
1.14.48
Steps to reproduce
- Ask it to create some file.
- Create a new session and ask it to update the file
Screenshot and/or share link
Operating System
Windows 11
Terminal
Windows Terminal
Description
Same as #10360, reportedly fixed. Still/again happening sometimes. I'm not sure when.
Some of the reported problems by the LLM (GPT-5.1):
Maybe the model just produces illegal patches?
Plugins
None
OpenCode version
1.14.48
Steps to reproduce
Screenshot and/or share link
Operating System
Windows 11
Terminal
Windows Terminal