umputun / agterm / 33692913491

97%

push

github

fix(control): honour the pane aliases session.type, session.text and font accept (#530)

`.claude/rules/control-api.md:251` says `session.type --pane` accepts `primary|left|top`,
`split|right|bottom` or `scratch`, and the shared CLI `validatePaneArgument` does accept all of them.
The app then matched raw spellings, so an alias passed `validate()` on the client and was refused by
the server. On 0.26.0 `agtermctl session text --pane split` answers `invalid pane: split`, the same
for `primary` and `bottom`, on `session.type`, `session.text` and `font.inc/dec/reset`.
`session.status` and `session.restore` were never affected: they already parse in the dispatcher.

The spelling is parsed once now, in the dispatcher, and the host actions take a `StatusPane`. Three
consequences past the aliases resolving: an unknown pane is unrepresentable in the app target, which
is where the module boundary wants the parsing, and three hand-rolled switches lose their
`invalid pane` arm; the aliases resolve for a raw socket client too, not only behind the CLI; and the
three `font.*` arms share one parse rather than three copies of the argument.

Nothing else changes on the wire. These three keep the `invalid pane: <value>` rejection they have
answered since they shipped (#46, #90, #188), which the reference records as their difference from
`session.status`'s pinned string, so `parseSurfacePane` carries that wording while `parsePane` keeps
the pinned one. Unifying the two is a separate decision from parsing the value.

`session.text` keeps checking the extent before the pane, so `--all --lines` still reports the extent
error first; a test pins that ordering.

31 of 31 new or added lines in 2 files covered. (100.0%)

9625 of 9877 relevant lines covered (97.45%)

4131358.23 hits per line