You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Files on Windows use \r\n as the line ending whereas POSIX systems use \n.
This poses a problem when editing Windows files from WSL - in the fs_read tool, we only send the content joined by \n. This causes context-mismatch issues with the model whenever it attempts to make changes, since it is not expecting line endings to contain a \r prefix.
Instead, we should detect what line endings to use (\r\n vs \n) when editing a file since it is possible to edit Windows based files that use \r\n while in WSL.
This is a great opportunity to enhance cross-platform compatibility. I'd be happy to investigate implementing a line ending detection mechanism. The approach could involve:
Detecting the existing line ending style in the file (\r\n vs \n)
Preserving that style when writing back changes via fs_write
Adding tests for both Windows and POSIX line ending scenarios
Would this align with the expected solution? I can start working on a PR if this direction makes sense.
I’m investigating a focused fix that preserves each file’s existing LF/CRLF style at the shared read/write boundary, with regression coverage for both endings. I found no linked implementation PR; I’ll keep the change scoped to this issue.
Activity
Scarmonit commentedon Oct 30, 2025
This is a great opportunity to enhance cross-platform compatibility. I'd be happy to investigate implementing a line ending detection mechanism. The approach could involve:
Would this align with the expected solution? I can start working on a PR if this direction makes sense.
voidstackloop commentedon Sep 13, 2026
I’m investigating a focused fix that preserves each file’s existing LF/CRLF style at the shared read/write boundary, with regression coverage for both endings. I found no linked implementation PR; I’ll keep the change scoped to this issue.