park's not-carried report counts a leg's unpushed commits against the root's tracking_branch, not the leg's own branch:, so work on a declared branch is never reported #242

What happens

park's not-carried report counts the commits a leg has on its tracking branch that origin has never seen, against the ROOT's tracking_branch, for every leg. The root's bootstrap.py places a leg on str(leg.get("branch") or branch), so a leg whose row in project.yaml declares its own branch: has its unpushed work on THAT branch, and park asks about the wrong one.

• park report_repos (park:526-548 at main f5c3bcd) emits each leg's row as leg<TAB><root's branch><TAB><path>, with the comment "A LEG'S branch is the ROOT's tracking_branch: — that is the one value scripts/bootstrap.py places every leg on" (park:536-537). That comment is not true for a leg that declares a branch:.
• not_parked_report (park:579) hands that branch to unpushed_on_tracking (park:559), which needs both refs/heads/<branch> and refs/remotes/origin/<branch> and counts origin/<branch>..<branch>.
• So for a leg with branch: develop, three local commits on develop that origin has not seen are never counted: park asks about main, finds nothing ahead, and the report ends without naming the leg. That report is what tells a person which work park did NOT carry, so it is data-loss-adjacent: the miss is silent.

resume and status had the same assumption and are addressed by #238 (for #235): resume's root_legs_ok and status's estate_repos now take each leg's branch from leg_branch_of, the shared helper that reads the leg row's own branch:, else the root's tracking_branch. park was left out because #235's finding does not name it.

Found by the independent review of #238 (review 5561239715).

Suggested fix

In report_repos, give each leg's row leg_branch_of "<root>" "<leg>", falling back to the root's branch on an empty answer, as status's estate_repos does in #238:

while IFS= read -r leg; do
    legbranch="$(leg_branch_of "$ESTATE_ROOT" "$leg")"
    printf 'leg\t%s\t%s\n' "${legbranch:-$branch}" "$ESTATE_ROOT/$leg"
done < <(submodule_paths "$ESTATE_ROOT")

(and the same in the family loop, with $ESTATE_DIR/$member as the root). leg_branch_of is already in park's copy of the shared resolver block, unused. Then replace the comment at park:536-537.

Add a test beside the not-carried report's: a root whose spec leg declares branch: develop, with local commits on develop that origin lacks; it must be reported, and it must fail without the change. The unpushed line prints the branch, so it can name develop.

🤖 Generated with Claude Code