docs(rust): protocol 13 — the reload as the rigs showed it (W1-W5, D180)

PROTOCOL.md §19.1: how the reload is asked for (Carbon: c.reload on the
server console), which log is read (Carbon.Core.log on Carbon), the pinned
failure lines on both frameworks, a load confirmed by a log read, and the
compiler's give-up retried once. PLAYER_WALK.md: the Carbon reload path and
failure words as walked. PLAN_FIXES.md: D180.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01E14m6SuuY6i1vASFeGDBeY
This commit is contained in:
2026-09-26 20:21:26 -05:00
parent cfa37c44e3
commit b8d9740ed0
3 changed files with 48 additions and 15 deletions

View File

@@ -1987,13 +1987,41 @@ depended on timing (F10).
- A second write while one is pending is still refused `busy`. The window is longer now, so `busy` is
more likely, and a website should hold its save button while it waits.
**Failing on evidence (D178).** While a reload is pending the plugin reads the framework's newest log
file once a second, from where it stood when the reload was asked for. A line that names the target as
a whole word (`Kits`, `Kits.cs`, `'Kits v4.4.9'`, never `KitsExtended`) beside one of *error while
compiling*, *failed to initialize*, *failed initializing*, *failed to compile*, *failed compiling*,
*failed to load*, *failed loading* or *unable to load* is a failure, and the edit is rolled back at
once. The bridge's own lines are skipped. A slow compile logs nothing until it succeeds, so it is simply
waited for. The Carbon pass of the re-walk pins Carbon's wording.
**How the reload is asked for.** On Oxide, through `Interface.Oxide.ReloadPlugin` (found by reflection),
falling back to `oxide.reload <name>` on the server console. **On Carbon, always `c.reload <name>` on the
server console** — the command an operator types at the panel. Carbon 2.0.259's Oxide-compatible
`ReloadPlugin` is not a reload: on the rig one call unloaded Kits and never compiled it again, and the
next did nothing at all (walk finding W3, 2026-09-26).
**Failing on evidence (D178).** While a reload is pending the plugin reads the framework's log once a
second, from where it stood when the reload was asked for. On Oxide that is the newest `*.txt` in its log
directory; **on Carbon it is `Carbon.Core.log`**, chosen by name — `carbon/logs` holds five `*.log`
files and no `*.txt`, and its Harmony log is the one written most often (W4). `rg.config` prints the
file it follows. A line that names the target as a whole word (`Kits`, `Kits.cs`, `'Kits v4.4.9'`,
never `KitsExtended`) beside one of *error while compiling*, *could not initialize*, *failed to
initialize*, *failed initializing*, *failed to compile*, *failed compiling*, *failed to load*, *failed
loading*, *failed iloadconfig* or *unable to load* is a failure, and the edit is rolled back at once. The
bridge's own lines are skipped. A slow compile logs nothing until it succeeds, so it is simply waited for.
Pinned on the rigs, for a config whose plugin's `LoadConfig` throws:
| Framework | The line | Then |
|---|---|---|
| Oxide | `[Error] Could not initialize plugin 'Kits v4.4.9' (JsonReaderException: …)` | "Previous version of plugin failed to load"; the plugin is **not** loaded |
| Carbon 2.0.259 | `[ERRO] [Kits] Failed ILoadConfig (Could not convert string to double: …)` | "Loaded plugin Kits"; the plugin **is** loaded, broken |
The first line carries the exception, so it is the one the outcome's `reason` quotes (W2).
**A load is confirmed by a log read, not by the hook (W5).** Carbon loads a plugin whose config it could
not read and raises `OnPluginLoaded` in the same instant it logs the failure. So the hook is noted, and
the job settles on the first log read at least a second after it: a failure found there rolls back; none
found means the reload worked. The cost is about a second on every save.
**The compiler giving up is not a failure of the file (W1).** Oxide shuts its compiler down after a
minute idle, and a reload asked for in that moment is dropped with `Timed out waiting for plugin to be
compiled: <name>` while the plugin runs on, loaded, on its old file. On that line the plugin asks for the
reload **once more** under a fresh ceiling (the compiler restarts on demand). A second such line settles
the job at once by what is loaded — the ceiling's judgement below, without waiting for the ceiling.
**The ceiling, 30 s per phase (F10).** A reload that says nothing either way is judged by **what is
loaded**, not by the clock:
@@ -2026,8 +2054,8 @@ plugin is back.
- **No outcome follows** if the bridge itself is unloaded while a write is pending. The files stay as
written; the website calls the write lost after its deadline and re-reads.
`rg.config` shows a pending write's target, phase (`edit` or `restore`) and the seconds spent of the
ceiling.
`rg.config` shows the log file the plugin follows (`log=`), and a pending write's target, phase (`edit`
or `restore`) and the seconds spent of the ceiling.
### 19.2 The sidecar