starred/orca
1
0
Fork 0
mirror of https://github.com/stablyai/orca.git synced 2026-09-26 16:02:43 +00:00
Code Issues Packages Projects Releases Wiki Activity

docs(mobile): wrap the recording README at the width the rest of it uses

Browse Source 
Seven lines this branch added ran past 100 columns, worst 124. No wording changed.
Markdown is outside `recorderSha256`, so no golden moves.

Claude-Session: https://claude.ai/code/session_01JNnE9qzUZMMnqpZWCqM3nb
This commit is contained in:
Jinwoo-H
2026-09-14 14:15:11 -04:00
parent 3b3e660776
commit 95b81968cd
1 changed files with 16 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
mobile/src/test-support/rpc-recording/README.md
+16 -10
View File
@@ -60,7 +60,8 @@ visible, it does not make the reduction itself observable.
## Golden schema
Each file records `runnerVersion`, `baseline`, `lockfileSha256` (mobile's lockfile),
`recorderSha256`, `adapterSha256`, `scenarioSha256`, `platform`, `scenarioVersion`, `projectionVersion`,
`recorderSha256`, `adapterSha256`, `scenarioSha256`, `platform`, `scenarioVersion`,
`projectionVersion`,
`goldenFormatVersion`, `operation`, `family`, and `namedDeltas`. `platform` and `lockfileSha256`
are provenance and are not compared: a dependency or OS that changes behaviour changes the trace
itself, so comparing them would only fail candidates on unrelated bumps. The rest are pinned.
@@ -81,7 +82,8 @@ against every domain branch in flight.
`mutants/` is excluded for a different reason: nothing there is pinned by anything. A file that
cannot change a recording is not provenance for one, and pinning it would claim a provenance the
golden does not have — while charging every domain that adds a mutant a re-record of all 153 files.
golden does not have — while charging every domain that adds a mutant a re-record of all 153
files.
The mutant table, the per-family mutant registry, the reference states and the suites that apply
them all live there. What makes the exclusion sound is that no recording can reach them: the loader
takes a resolved mutation spec instead of importing a table by name, so nothing on the recording
@@ -112,7 +114,8 @@ it. The manifest used to be an input to `recorderSha256` instead, which made eve
a function of every other family's scenarios: adding one domain's family re-digested all 153 files
and put a conflict on that line in every domain branch in flight. Which goldens a manifest derives
lives in `derived-goldens.ts`, so the digest is a function of the same derivation that records the
file rather than of a restatement of it; `golden-header-digest.test.ts` pins what both separations buy.
file rather than of a restatement of it; `golden-header-digest.test.ts` pins what both separations
buy.
Checkpoints contain ordered sender calls and serialized physical application payloads, action and
request settlements, projected state, and ordered external effects. Sender args have three
@@ -237,7 +240,8 @@ ORCA_BACKGROUND_LAUNCH=1 pnpm --dir mobile test src/test-support/rpc-recording
```
Mutants are the defect evidence. `mutants/operation-mutations.ts` holds one anchored
source edit per adapter family, and every family's recording must change visible state when its mutant is applied,
source edit per adapter family, and every family's recording must change visible state when its
mutant is applied,
which is what shows that family's `state()` projection observes the operation's real output.
Anchors are asserted to match exactly one site, because a repeated anchor would half-apply while
still counting as applied. Mutants replace the expression in memory, then run the same real hook.
@@ -280,7 +284,8 @@ It is not a substitute for reading the diff. Three facts bound it, all learned t
reply kills it on five matrix goldens. The lesson is about the skip, not about that call site: a
generator that opts a family out without failing is indistinguishable from coverage.
`mutants/probe-hole-witness.test.ts` closes the first two and keeps them closed. It asserts the hole and the closure
`mutants/probe-hole-witness.test.ts` closes the first two and keeps them closed. It asserts the hole
and the closure
together: each probe must kill its mutation _and_ every pre-probe scenario of the same operation
must still survive it. A probe that stops being load-bearing fails instead of lingering.
@@ -331,12 +336,13 @@ copied in, `node_modules` symlinked, `RPC_FOUNDATION_GOLDENS` pointed at a scrat
copy the result back and run the candidate suite here. Format the recorder before recording: an
`oxfmt` pass afterwards moves `recorderSha256` again. A recorder-only branch that has merged main
is not the awkward case: its product tree is main's, so repin `baseline` to main's tip and record
in place — there is no migrated source for the goldens to be recorded against. Adding or editing one domain's module under
`adapters/` no longer needs any of this: only that domain's goldens move, and they re-record from
its own branch like any other behaviour change. Adding a mutant, a probe or a suite that does not
record needs none of it either, and moves no golden at all.
in place — there is no migrated source for the goldens to be recorded against. Adding or editing
one domain's module under `adapters/` no longer needs any of this: only that domain's goldens move,
and they re-record from its own branch like any other behaviour change. Adding a mutant, a probe or
a suite that does not record needs none of it either, and moves no golden at all.
If your call site carries a mutation anchor in `mutants/operation-mutations.ts`, rewriting it will make the
If your call site carries a mutation anchor in `mutants/operation-mutations.ts`, rewriting it will
make the
anchor match zero sites. Re-anchor the same defect at its new home rather than deleting the mutant:
#20499 broke five anchors that way, and each one had a new home.