Levi Neuwirth
5cc1a83583
CI caught this on all four test legs. The forwarder body called `pmacs.command.invoke_interactive`, which raises when the alias is reached through `pmacs.command.invoke` — and `tests/config_registry_acceptance.rs` does exactly that, three times. The acceptance pin passed throughout because it drives the M-x path, which is the path the framing spent three review rounds getting right. Being right about one entry point is not the same as covering the command, and a rename touches every caller of the old name regardless of how it is reached. Plain `invoke` is also the correct semantics rather than merely the working one: the interactive-command boundary is rotated once, by whatever entry point the user actually used, for the name they actually typed. Rotating again on the inner call would record a second boundary for a command the user never invoked. Adds `d8c`, which invokes both forwarders programmatically. Bitten by restoring `invoke_interactive`: the new pin fails alongside the three config-registry tests that found it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Lv428Fth9LRtffwJSsqH7T |
||
|---|---|---|
| .. | ||
| api | ||
| commands | ||
| hooks | ||
| keymaps | ||
| menus | ||
| packages/repl | ||
| queries/latex | ||
| runtime | ||