mirror of
https://github.com/l0ng-ai/tty7.git
synced 2026-09-22 00:02:23 +00:00
l0ng-ai
573064064e
This branch settles a repository's key on one spelling, but it settles it with *this* machine's rules everywhere: `local_spelling` folds `/` to `\` and strips `\\?\` on any Windows build, whatever machine the path is on. Every boundary it was applied at also serves a remote workspace. - `core::git::git_path` parses `rev-parse` output client-side, for `probe`, `probe_status`, `probe_diff` and `worktree`, whichever host ran the command; - `GitStatusCache::key` is `ByHost`, and `refresh_git_status` probes a remote pane's cwd under that pane's `HostId`; - `scm_repo_root` resolves the root every SCM write then runs `git` from. So on a Windows client with an SSH or WSL workspace `/home/u/src` became `\home\u\src`, and that string is what goes back over the wire verbatim — `RemoteHost`'s `wire_path` is `to_string_lossy` — as the cwd of the next `git`, as the `ScmData` key and as the `.git` watch path. It names nothing on the far side, and it is what the sidebar draws. `path_spelling::spelling_on` puts the question where #795 puts it: path syntax is a property of the machine the path is *on*, not of the one asking. A local host still gets the whole rule, so the `\\?\` fix and the editor-watch fix are untouched; a remote host is handed back exactly what it said, which is what this tree did before the rule existed. Two regression tests, both ungated because Windows is the only client they can fail on. `git_status::a_repository_on_another_machine_keeps_that_ machines_spelling` fails without this change with `left: Some("\\home\\u\\src") right: Some("/home/u/src")`.