starred/orca
1
0
Fork 0
mirror of https://github.com/stablyai/orca.git synced 2026-09-22 00:02:31 +00:00
Code Issues Packages Projects Releases Wiki Activity

docs(mobile): refresh the recorder corpus counts

Browse Source 
397 scenarios, 787 goldens, 790 tests from the README's own three-file command.
The 44 salvage goldens are unchanged; only the total they are quoted against
moved.

Claude-Session: https://claude.ai/code/session_01JNnE9qzUZMMnqpZWCqM3nb
This commit is contained in:
Jinwoo-H
2026-09-18 00:21:45 -04:00
parent cf5aa5c391
commit c519c2027d
1 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files
mobile/src/test-support/rpc-recording/README.md
+4 -4
View File
@@ -183,7 +183,7 @@ off the context `client-context.tsx` keeps module-private, and each used to carr
`exports.recorderHostClientContext = Ctx;`. That string names a local no type checker follows, so
five spellings were five independent ways to reach a `ReferenceError` seconds into a recording.
`hostClientContextExposure` is the one copy; the trade is that it sits inside `recorderSha256`, so
editing it re-records all 778 goldens rather than the five families. A rename of the local is still
editing it re-records all 787 goldens rather than the five families. A rename of the local is still
invisible to `tsc` — nothing short of editing the product module makes a private local checkable —
so `adapter-seam.test.ts` asserts the declaration it names exists exactly once, and refuses a sixth
inline copy.
@@ -378,8 +378,8 @@ families because no reference states are defined for them.
## What this oracle does and does not see
It replays 393 manifest scenarios against frozen goldens and fails on any divergence: 778 goldens
over 781 tests, all inside `pnpm --dir mobile test`. Counted with
It replays 397 manifest scenarios against frozen goldens and fails on any divergence: 787 goldens
over 790 tests, all inside `pnpm --dir mobile test`. Counted with
`python3 -c "import json;print(len(json.load(open('mobile/rpc-foundation/pilot-scenarios.json'))['scenarios']))"`,
`find mobile/rpc-foundation/goldens -type f | wc -l`, and the reported total of
`vitest run src/test-support/rpc-recording/{pilot,family}-recordings.test.ts src/test-support/rpc-recording/derived-goldens.test.ts`. Counts quoted further down are measurements of
@@ -613,7 +613,7 @@ the drop happened under, and records a non-empty report as a `reply-salvage` eff
operation, the method, the decoded variant, the dropped paths and the count. Nothing in the product
tree changes: the report was already being built and thrown away.
44 of the 778 goldens carry one, and every other checked read in the corpus decodes its reply
44 of the 787 goldens carry one, and every other checked read in the corpus decodes its reply
whole (`grep -l reply-salvage mobile/rpc-foundation/goldens/*.json | wc -l`). The matrix varies the
envelope a host sends rather than the shape of a row inside a result, so on most families this
observation pins an absence rather than a recorded drop. What it buys is the next tightening: an element or member schema narrowed so a recorded row stops parsing moves the