xiongyuanwen-byted 3558e5b548 fix(sheets): let every command that resolves a sheet also defer asking for one ...

Review found the selector deferral half-wired. The requirement is easy to
satisfy in Execute and easy to miss in Validate, because the two read the
selector from different places: Execute passes the RESOLVED pair into the
input builder, Validate passes the raw flags. A builder calling
requireSheetSelector therefore passes for one caller and fails for the other,
with nothing tying them together.

Sweeping every command whose Execute calls resolveSheetSelectorExec found nine
still rejecting in pre-flight -- more than the review named: +cells-merge,
+cells-unmerge, +rows-resize, +cols-resize, +range-copy, +range-move,
+dim-delete (both --range and --ranges), +dim-hide / +dim-unhide,
+dim-group / +dim-ungroup, and the float-image writes. For those the resolver
their Execute calls was unreachable.

TestSheetSelector_DeferralReachesEveryResolver is that missing tie, and its
counterpart pins that --dry-run keeps the rejection: a preview sends nothing,
so it cannot ask the workbook either, and printing a request whose sheet the
CLI would have filled in would show a call that never existed.

Also asserts the typed envelope in the sheet-prefix conflict E2E case rather
than the message text alone. type / subtype / param are what a caller branches
on; the prose is free to be reworded, and pinning only the prose is what made
that test fail on a reworded message once already.

2026-09-14 15:27:39 +08:00

..

cli_e2e

fix(sheets): let every command that resolves a sheet also defer asking for one

2026-09-14 15:27:39 +08:00

plugin_e2e

feat: add lazy API catalog routing (#2232)

2026-09-10 13:29:26 +08:00

sidecar_e2e

feat: add lazy API catalog routing (#2232)

2026-09-10 13:29:26 +08:00