Thursday, October 29, 2026 · Ubuntu 24.04.5 LTS (kernel 6.12.94)

Dogfood on linux-cloud: Newcomer

6 repos · 17 commands · 3 broken · 1 slow · 5 filed

Newcomer on linux-cloud: installed release 0.5.24 with the website script, built main e84d7c9df, walked first-run, beginner mistakes, conflict handling, history/change-id behaviour, teammate push/pull, snapshot checkout (checkout/checkin/integrate against grit serve), a first-time library program, and re-checked the wave's fixes. Everyday paths and errors are clear and consistent across human/--json/--markdown; five issues filed (one real state-machine bug in rebase, one library API trap, three small UX gaps).

os: Ubuntu 24.04.5 LTS (kernel 6.12.94) · x86_64
machine: Cursor cloud VM, 4 vCPU Intel Xeon, 16 GB RAM
versions: grit 0.5.24 · main e84d7c9d · 2.43.0

repositories6

repo	files	commits	.git	notes
rust-lang/cargo	3,078	23,596	85 MB
octocat/Hello-World	1	3	1 MB
scratch/30k-files	30,000	1	5 MB	200 dirs x 150 files, grit-committed
scratch/submodule-super	2	5	1 MB	one submodule with diverging gitlink bumps
scratch/partial-clone-src	6	3	1 MB
scratch/small repos s1-s7, r1, rx, cf, cf3, h1, m1, n1	—	—	—	beginner-mistake and conflict scenarios

commandsgrit vs. system git, wall time

command	repo	grit	git	grit/git	result	notes
grit status	rust-lang/cargo	15 ms	11 ms	1.3×	ok
grit log -n 100 --oneline	rust-lang/cargo	3.7 ms	2.4 ms	1.5×	ok
grit blame Cargo.toml	rust-lang/cargo	86 ms	112 ms	0.8×	ok
grit diff HEAD~500 HEAD	rust-lang/cargo	231 ms	251 ms	0.9×	ok
grit switch old / master (3000 commits apart)	rust-lang/cargo	186 ms	175 ms	1.1×	ok
grit log -n 20 -- <path>	rust-lang/cargo	5.7 ms	3.9 ms	1.5×	ok
grit branch -a	rust-lang/cargo	2.3 ms	1.7 ms	1.4×	ok
grit stash (200 modified files)	rust-lang/cargo	113 ms	—	—	ok	stash pop 47 ms vs git 44 ms
grit clone https://github.com/rust-lang/cargo	rust-lang/cargo	2.41 s	4.37 s	0.6×	memory	2.6x git peak RSS, faster wall; fsck clean; not filed (clone memory covered by existing perf work)
grit commit -m import (30k new files)	scratch/30k-files	684 ms	986 ms	0.7×	ok	git figure is add -A + commit
grit status	scratch/30k-files	43 ms	24 ms	1.8×	ok
grit gc	scratch/30k-files	35 ms	—	—	ok
grit checkout <url> --path src/sources	rust-lang/cargo	89 ms	—	—	ok	42 files, 196 KiB download from grit serve
grit clone --blobs none <plain repo>	scratch/partial-clone-src	65 ms	—	—	ok	warns that the server ignored the filter
grit rebase --continue after grit commit	scratch/small	—	—	—	wrong	#2293
Repository::init + commit_worktree with captured environment	scratch/lib program	—	—	—	wrong	#2294
grit reset --hard + hinted recovery	scratch/small	—	—	—	wrong	#2296

worked7

›Install script and `grit update` work; release 0.5.24 runs from a clean HOME with a clear identity-unknown message
›Typo suggestions, git-habit hints (init -q, checkout -b, clean -d, commit --no-edit) and structured errors with kind/hint in --json and --markdown
›commit --staged, commit -- paths, commit.autoStage=false, merge --check, -X, branch --force-delete all behave as decided
›change-id header kept across amend/rebase/history and fresh on pick/revert; git fsck clean everywhere
›Teammate push/pull and grit serve snapshot checkout/checkin/integrate end to end, with actionable conflict messages
›Library program with typed errors produced a git-fsck-clean history using only public API
›30k-file commit packs the objects and beats git add+commit; cargo-size read paths are within 1.5x of git

problems13

›#2293 grit commit during a rebase conflict then rebase --continue fails (filed)
›#2294 Repository::init ignores InitOptions::environment for the returned handle; failed identity check leaves dangling tree (filed)
›#2295 misleading mid-rebase errors for pull/merge/pick/rebase (filed)
›#2296 reset --hard recovery hint is a mixed reset (filed)
›#2297 -n on clean/add/rm/mv gets clap's value tip instead of --dry-run (filed)
›`grit` with no args outside a repo: bare 'not a git repository', no init/clone pointer (not filed, minor)
›merge --continue auto-stages while pick/rebase --continue need grit add (documented, inconsistent with decision 2; not filed)
›Empty-blob dangling object (e69de29) left by rebase/history conflict paths (harmless, not filed)
›grit update re-downloads when already latest without saying so (not filed, minor)
›`grit status` on main says 'even with main' when on the target branch itself (confusing for newcomers, not filed)
›Non-executable hook silently ignored; hook failure has no --no-verify hint (not filed, minor)
›docs library/merge.md claims merge orchestration stays in grit-cli while integrate_branch implements it (not filed)
›grit-lib cannot be built with the machine's default cargo 1.83: edition2024 manifests fail before the MSRV message (inherent to cargo, noted under #2227)

bugs filed5

#2293 #2294 #2295 #2296 #2297already reported: #2155 #1909 #1814 #2154

next time

›Real GitHub account: grit auth device flow and a first push (still impossible here)
›Case-colliding paths on macOS for #2226 (not checkable on Linux)
›merge/pick/rebase --continue consistency (auto-stage) once the maintainer decides; retest `grit commit` mid-rebase after #2293 is fixed, including multi-commit rebases and --exec
›grit serve at scale with several concurrent checkouts/checkins, plus slices (grit slice new/edit) and integrate races
›Library: custom CredentialProvider over Remote::fetch, RefStore/ObjectStore embedding, and a program written only from docs.rs pages (not the in-repo examples)
›Large-repo clone memory (cargo 2.6x git RSS) and git/git scale for checkout/pull