kettle-dev / kettle-family / 37934679613

88%

push

github

🔥 Reject --start-at for wave-dispatched commands

The previous commit fixed the resume hint that template and test emit, but
left the underlying hazard reachable: --start-at passed by hand, or
replayed from shell history, still selected positionally and silently
dropped members that never ran.

Selection#select_start_at drops everything before the named member. For
sequentially dispatched commands that is correct, because they abort at the
first failure so the pending set really is a suffix. For template and test it
is not: they dispatch in dependency waves and a concurrent queue, where a
failure only sets a stop flag, so pending members are interleaved through the
order.

Selection now rejects the flag for those commands and exits 1 with a message
naming the equivalent --only selection, following the existing
validate_release_state_only_filter! class-method convention so the CLI does
not have to know which commands are affected.

    $ kettle-family template --start-at nomono
    kettle-family: --start-at is not supported for template: members are
    dispatched in dependency waves, so a positional resume would silently
    skip members that never ran. Use --only nomono (or the resume hint from
    the previous report) instead.

The dispatch classification now lives once, on Selection, because that is
where the flag is rejected. Report delegates to it instead of keeping a
second copy of the same constant, which also removes a duplicate method
definition that had been introduced in the previous commit.

release deliberately keeps accepting --start-at. It also dispatches in waves
(release_waves, whose order differs from the Orderer order -- measured:
kettle-wash moves 10 to 6 and kettle-drift 7 to 12), but it uses the flag for
MEMBER@BRANCH branch-lane selection through release_target_branches, and
existing specs exercise `release --start-at alpha@r1_9-even-v2`. Its resume
hints already use --only and never emit a positional hint, so there is no
silent... (continued)

2755 of 3537 branches covered (77.89%)

Branch coverage included in aggregate %.

10 of 10 new or added lines in 3 files covered. (100.0%)

6778 of 7265 relevant lines covered (93.3%)

109.99 hits per line