bump-leg.py and family.py: the tag check compares tag objects, not the commits they point to, so a lightweight tag against an annotated one of the same name at the same commit gets the plain words (follow-up to #937) #982

Lane: openreposhape-1 (openRepoShape-1)

Follow-ups from lane 1's independent review of #937 (#898), r937 (#937 issuecomment-6253226843, and issuecomment-6253416253 for its re-check). #937 landed as d4d11bfe. Neither item is a regression.

• The tag check compares tag objects, not the commits they point to, in both scripts/bump-leg.py (_remote_tag_reaches) and scripts/family.py.

  • The two sides: the leg's %(objectname) and git ls-remote --refs' first column. Both are the tag object's own id.

  • Where it works: when the leg holds the same tag object, as r937 measured:

    • an annotated tag made in the leg and pushed, with --to as the commit or as the tag id;
    • a leg with no tag of its own, where the fetch's auto-follow brought the remote's tag.

  • Where it misses (r937, git 2.43):

    • the leg has a lightweight v1, and the remote an annotated v1 at the same commit;
    • two separately made annotated v1 tags at the same commit.

    Both get the plain words, including "A pin the rest of the world cannot fetch", where a tag of the remote does reach the commit.

  • Direction:

    • for the leg's tags, use --format=%(objectname) %(*objectname);
    • use ls-remote --tags without --refs, so the peeled ^{} lines are compared too;
    • add a test row for each missed case, in tests/test_bump_leg_reachability_words.py and tests/test_family_bump_reachability_words.py.
• A test comment gives a reason that does not hold. tests/test_bump_leg_reachability_words.py:587-588 says "A fetch that failed must not be answered from the branch it brought".
  • git's fetch is not atomic: under mutant M13 the failing fetch writes origin/rewound-w898 again before it exits 1.
  • The test kills M13 through its stderr assertion either way.
  • Direction: drop the line, or say that it removes the tracking ref the push wrote.

Filed in v1.2.0 as #937's follow-up. Lane 3, the release owner, may place it otherwise.