fix(git): preserve full status paths and untracked files

[em0t]

Summary

• switch default rtk git status to --porcelain -b -uall so dirty paths are no longer hidden behind grouped summaries, overflow markers, or directory-level collapse
• preserve RTK's status framing (* branch, clean — nothing to commit) while keeping every dirty path visible
• add focused regression coverage for default status behavior, explicit-args passthrough, rename/conflict preservation, and untracked directory expansion

Minimal Cases

rtk git status Hidden Dirty Paths

Previous RTK (develop)

* main
~ Modified: 17 files
   src/file01.rs
   src/file02.rs
   ...
   src/file15.rs
   ... +2 more

This PR

* main
 M src/file01.rs
 M src/file02.rs
 ...
 M src/file15.rs
 M src/file16.rs
 M src/file17.rs

Untracked Directory Expansion

Raw

On branch main
Untracked files:
  (use "git add <file>..." to include in what will be committed)
	tmp/

nothing added to commit but untracked files present (use "git add" to track)

Previous RTK (develop)

* main
? Untracked: 1 files
   tmp/

This PR

* main
?? tmp/c.txt
?? tmp/nested/d.txt

Testing

• cargo fmt --all
• cargo clippy --all-targets
• cargo test --all

[CLAassistant]

All committers have signed the CLA.

[em0t]

Could a maintainer please approve and run workflows so CI can start for review?

[aeppling]

Hello @em0t

Thanks for those contributions !

I'll review this today , currently focusing on bugs and filters quality.

Did you check if there was any issues related to your 4 PRs ?

[em0t]

Thanks @aeppling. I didn’t find existing upstream issues specifically tracking these 4 cases.

These 4 PRs came from an internal local A/B evaluation we ran while assessing RTK for rollout in our team. We used Claude Code + Codex, wrote up a report, and these were the main issues we found. I split them into 4 small PRs by issue category to keep review clearer and each fix easier to reason about.

Also yes, all 4 PRs include automated regression tests for the behavior they change, and I ran cargo fmt --all, cargo clippy --all-targets, and cargo test --all locally on each branch.

To be transparent, I haven’t manually reproduced every case end-to-end by hand yet.

[aeppling]

Ok thanks, passing by this i see the output of git status is :

Modified files:
  src/a.rs
  src/b.rs
Untracked files:
  tmp/c.txt

But actually, isn't "modified" and "untracked" enough for LLM to understand and compress a bit more ?
Since we are removing some filters part, would be interesting to add this if it make sense for you so we still get nice compression globally on git commands

[em0t]

Yes, that makes sense. I’ll follow up with a smaller optimization for the git status formatting afterwards.

[aeppling]

Thank you !

[em0t]

@aeppling Follow-up pushed in this PR. Default git status now keeps every dirty path visible but switches the default output to porcelain-like status lines (##, M , ??) instead of grouped labels, so it compresses better without hiding files. I also removed the now-dead status limit config/docs and re-ran cargo fmt --all --check, cargo clippy --all-targets, and cargo test --all locally; clippy still shows the same pre-existing warnings outside this diff.

[em0t]

Thanks for the review. I narrowed this PR further to keep the scope as small as possible.

Changes since the last round:

• kept the scope in src/cmds/git/* only; no core config changes
• preserved the existing RTK status framing (* branch, clean — nothing to commit)
• kept only the behavior fixes:
  • default rtk git log no longer hides merge commits
  • default rtk git status no longer hides dirty paths and now expands untracked directories via -uall
• reduced the regression coverage to the focused cases for these behaviors
• updated the PR description to match the current behavior

Would appreciate another look when you have a moment. If anything still feels broader than it should be, I can trim it further.

[em0t]

@aeppling I thought about it more and removed the git log merge-commit change from this PR.

What’s left is just the two git status correctness fixes: preserving all dirty paths and expanding untracked directories.

Would appreciate another review when you have a chance.

[aeppling]

Hey @em0t

Additional thoughts , you bypass the porcelain flag addition when user run a "git status" cmd without arguments.
But in case user run "git branch -b", we could also catch to still gain token (if git status -b for eg.) Since we add -b in flags that could make sense to catch as well for compress.