Practical Picks PP-268: earlier hands-on check record, October 1, 2026 Observed 2026-10-01 UTC, Linux x86-64. No remote push or public repository. Historical record. This small check ran on October 1, 2026, before Practical Picks adopted its research-only method on October 3, 2026. It is kept as dated, bounded evidence. It is not current testing, and it did not decide the article's recommendations, which since October 19, 2026 rest on public sources. Correction, October 19, 2026: upstream Git 2.43.0 falls within the affected range of CVE-2025-48384 (fixed upstream in 2.50.1 and later, and in patch releases 2.43.7, 2.44.4, 2.45.4, 2.46.4, 2.47.3, 2.48.2 and 2.49.1). Linux distributions often backport such fixes without changing the version number; for example, Ubuntu 24.04 fixed its Git 2.43.0 package in July 2025 (USN-7626-1). This record did not capture the package build of the host Git, so it does not establish whether that binary was affected. Statements below that call the host binary "affected" should be read in that light. This is a condensed record of exact fixture bytes, observed commands and checks. It is not a portable one-command reproduction package: the research repository retains the runnable scripts and task-local binary paths. Version-control command syntax is given below so a reader can repeat the steps in disposable repositories with those tools installed. Reader task: revise an earlier small-project change while keeping a later change, reverse one deliberate mistake, and hand the final repository to a separate local reader. Frozen files (literal contents, each line ends with LF): base.txt: Port: Aster / Ships: 3 revision-1.txt: Port: Aster / Ships: 4 revision-2.txt: Port: Aster / Ships: 4 / Weather: clear revised-1.txt: Port: Aster / Ships: 5 expected-final: Port: Aster / Ships: 5 / Weather: clear The slash in this summary separates lines; it was not present in the files. Frozen source SHA-256: base 91fb6b01ad4f8e26f2612b5ab38ae1fc948dbfb52a04b1b08fc2013534738844 revision 1 718561889f7d676c750d742974c91d52c52628e549e2e18b535f4f7fabe1375b revision 2 7c28afc90b35ab9ed51872b226b99ee7e82475ebe2f37f1a033916ec508d476f revised 1 3bd8f96493764d9adaab55987c6e6487ec5702492c0d7f3830ec84e6b067b594 expected final 964b5922760c8f8f122da243b96dd8a9275c6ad6c4fe1e9d4a332c947aa5e103 Versions: host Git 2.43.0 for the first smoke; isolated Mercurial 7.2.4; official Jujutsu 0.45.1 Linux-musl release. Git's GHSA-vwqx-4fm8-6qc9 / CVE-2025-48384 advisory identifies host 2.43.0 as affected under crafted malicious-submodule checkout conditions: https://github.com/git/git/security/advisories/GHSA-vwqx-4fm8-6qc9 . The synthetic test contained no submodule or untrusted checkout. Git 2.43.0 is retained only as test provenance, not a current installation recommendation. Git's official install page listed 2.56.0 as latest on October 1, 2026: https://git-scm.com/install/ . The Jujutsu archive SHA-256 was f35438350b5d61963aac5dd74ede510b31d6b9690769d1a6268cf058cc825f72. Test repositories were separate task-owned directories. Git: git init -b main; create three commits from base, revision 1 and revision 2; GIT_SEQUENCE_EDITOR modified the first rebase action to edit; git rebase -i HEAD~2. Copy revised-1.txt over notes.txt, git add, git commit --amend --no-edit, then git rebase --continue. Replay stopped at an adjacent-line conflict. Copy expected-final.txt over notes.txt, git add, then continue. A deliberate overwrite was committed and git revert HEAD --no-edit made an inverse commit. A bare clone and a second working clone received expected-final bytes. Mercurial: hg init; create three changesets from the same files; enable the shipped histedit extension and use hg histedit -r 1 with first action set to edit. Copy revised-1.txt, continue; replay stopped at an adjacent-line conflict. Copy expected-final.txt, hg resolve -m notes.txt, continue. A deliberate overwrite was committed, then hg backout -r . --no-commit and a new inverse changeset restored it. A separate hg clone received expected-final bytes. Jujutsu: jj git init a fresh Git-backed workspace; create three changes using jj describe and jj new. jj edit @- changed the earlier revision, after which copy revised-1.txt. Returning to the descendant showed an explicit conflict; copying expected-final.txt over notes.txt and jj status snapshotted the resolution. jj restore --from @- notes.txt deliberately removed the weather line; jj op revert restored it. Only after conflict resolution did jj bookmark create main -r @ allow an ordinary git clone -b main of the colocated repository to receive expected-final bytes. No remote push was attempted. All final working files and three recipient copies matched expected-final.txt with diff -u exit 0 and SHA-256 964b5922760c8f8f122da243b96dd8a9275c6ad6c4fe1e9d4a332c947aa5e103. All three conflict resolutions used the predeclared expected bytes rather than claiming automatic conflict resolution. Git and Hg commit subjects retained the old phrase 'ships four' after the content was revised to five; this is a test-history hygiene limit, not a product result. Full-research follow-up, separate-file edit: config-base.txt: Port: Aster / Ships: 3 config-first.txt: Port: Aster / Ships: 4 config-revised.txt: Port: Aster / Ships: 5 weather-later.txt: Weather: clear Again, slash separates LF-terminated lines here; it is not file content. The base and first config files have the same bytes as the original base.txt and revision-1.txt. The revised config file has the same bytes as revised-1.txt. weather-later.txt SHA-256 is 5fc68dd08812e3dc960e3aad00401938ae76365c8a9355f3204eadf0ee702e70. For each CLI, repeat the prior init/base/earlier/later sequence with config.txt as the earlier file and weather.txt as the later newly added file. Change the earlier revision from Ships: 4 to Ships: 5 with git rebase -i HEAD~2, hg histedit -r 1, or jj edit @-. Continue or return to the descendant. No replay conflict appeared in any of the three fresh runs. diff -u against both frozen final files returned 0. Final config.txt SHA-256 was 3bd8f96493764d9adaab55987c6e6487ec5702492c0d7f3830ec84e6b067b594 and weather.txt was 5fc68dd08812e3dc960e3aad00401938ae76365c8a9355f3204eadf0ee702e70 for each CLI. The private research record contains exact runnable scripts and the fresh run ID full-20261001-b. Additional Git recovery: a separate task-local build from the official Git 2.56.0 source archive (cached tar.xz SHA-256 26c56c296b38c0695b26fa95f475f1d01704d2d38e73465ca30b0b2f5dc789d3) reported git version 2.56.0. Optional network-related build features were omitted. The run completed the separate-file edit, saved its main tip, detached HEAD, and moved main back with git branch -f main HEAD~1. git reflog show main -2 showed the prior tip at main@{1}. git branch -f main main@{1}, followed by git switch main, restored the saved tip. A fresh git clone received both expected files. This tests one recent local branch-tip recovery, not expired or garbage-collected reflog data. It does not verify network operations or test the CVE exploit or fix. The private runnable script is run-git-current.sh; its fresh run ID was current-20261001-b. Unresolved Jujutsu-to-Git handoff: a separate Jujutsu 0.45.1 run recreated the adjacent-line conflict. jj status named notes.txt a 2-sided conflict; a bookmark pointed to that unresolved revision. An ordinary git clone -b unresolved received .jjconflict-base-0/notes.txt, .jjconflict-side-0/notes.txt, .jjconflict-side-1/notes.txt, JJ-CONFLICT-README and notes.txt with only Port: Aster and Ships: 5. The later Weather: clear line was absent from that notes.txt. This is not a clean handoff. The private runnable script is run-jj-conflict-handoff.sh; its verified fresh run ID was conflict-20261001-c. Jujutsu's official Git compatibility documentation explains this special conflict representation: https://docs.jj-vcs.dev/latest/git-compatibility/ . Resolve conflicts and verify the Git recipient before handoff. No speed, reliability-rate, security, network sharing, hostile-input, other-OS, large-repository, submodule, Git LFS, hook, or concurrent-writer inference follows from these synthetic local tests. The host Git 2.43.0 binary is affected by the cited advisory under conditions absent from the test; use a current patched Git release for real work. Jujutsu's resolved Git handoff does not establish every Git feature. Official documentation linked in the article supports documented features; direct observations are limited to this protocol.