Skip to content

Commit 8f4a7b7

Browse files
committed
test(coverage): keep macOS test-suite evidence as owed work and audit accepted reasons
The CI macOS replay lane runs agent-device test with --report-junit and gates on its exit status, so the test row is closeable with real lane evidence and stays known-gap on #1916. Each accepted-gap reason now states the absence it claims against the repository's lanes, and the gate's docstring encodes that rule.
1 parent 2f9c56f commit 8f4a7b7

2 files changed

Lines changed: 10 additions & 11 deletions

File tree

‎test/integration/command-coverage/declarations.ts‎

Lines changed: 7 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -74,7 +74,7 @@ const COMMAND_COVERAGE_DECLARATIONS = {
7474
),
7575
linux: linux.accepted(
7676
'No Linux live command creates a downloadable daemon artifact for inventory',
77-
'The Linux lanes run the local CLI, where every command output is a caller-owned local path; the daemon only inventories artifacts carried over remote artifact transport, so a Linux lane cannot create a downloadable inventory entry without a local-transport product decision.',
77+
'The Linux lanes run the local CLI, where command outputs are caller-owned local paths; request finalization inventories a response artifact only when the request carries a client artifact path (remote transport) or it is the sparse-snapshot fallback screenshot, which no Linux lane observes, so a Linux lane cannot produce a downloadable inventory entry without a local-transport product decision.',
7878
),
7979
},
8080
[C.devices]: {
@@ -142,7 +142,7 @@ const COMMAND_COVERAGE_DECLARATIONS = {
142142
),
143143
macos: macos.accepted(
144144
'No command-specific macOS doctor evidence exists',
145-
"The doctor checks are daemon-owned host diagnostics (version, state dir, inventory probes, Metro) with no macOS device branch, already pinned by the daemon's own doctor tests; a macOS lane would re-run that identical code, so macOS carries no command-specific doctor evidence.",
145+
'The doctor checks are daemon-owned host diagnostics — version, state dir, inventory probes, Metro — pinned platform-neutrally by the daemon doctor tests, and no macOS lane invokes the command: neither the System Settings replay nor the desktop and recording provider scenarios carry a doctor step.',
146146
),
147147
tvos: tvos.gap('No tvOS-specific doctor command evidence exists yet'),
148148
web: web.contract(
@@ -310,7 +310,7 @@ const COMMAND_COVERAGE_DECLARATIONS = {
310310
),
311311
macos: macos.accepted(
312312
'No command-specific macOS event timeline evidence exists',
313-
'The events timeline is daemon-owned session bookkeeping with no platform branch, already pinned for a daemon-backed session by request-router-events.test.ts; a macOS lane would re-run that identical code, so macOS carries no command-specific event timeline evidence.',
313+
'The events timeline is daemon-owned session bookkeeping, pinned by request-router-events.test.ts for a daemon-backed web session, and no macOS lane invokes the command: neither the System Settings replay nor the desktop and recording provider scenarios carry an events step.',
314314
),
315315
tvos: tvos.gap('No tvOS-specific session-event command evidence exists yet'),
316316
web: web.contract(
@@ -406,10 +406,7 @@ const COMMAND_COVERAGE_DECLARATIONS = {
406406
'full:fixture-replays',
407407
'fixture scripts run as a suite with JUnit artifacts',
408408
),
409-
macos: macos.accepted(
410-
'No command-specific macOS test-suite evidence exists',
411-
"The CI macOS replay lane dogfoods the test command to run the replay suite, whose per-command macOS claims are already carried by the replay:system-settings live rows; no lane asserts the test command's own suite result and JUnit output on macOS, and none will.",
412-
),
409+
macos: macos.gap('No command-specific macOS test-suite evidence exists yet'),
413410
tvos: tvos.gap('No tvOS-specific test-suite command evidence exists yet'),
414411
web: web.accepted(
415412
'No command-specific web test-suite evidence exists',
@@ -798,7 +795,7 @@ const COMMAND_COVERAGE_DECLARATIONS = {
798795
),
799796
macos: macos.accepted(
800797
'No command-specific macOS React Native inspection evidence exists',
801-
"react-native dismiss-overlay only answers meaningfully on a surface that carries a React Native overlay, and the repository's React Native fixtures run on iOS simulators and Android emulators; the macOS lanes drive desktop apps, so no macOS lane will cover this row.",
798+
"react-native dismiss-overlay only answers meaningfully on a surface that carries a React Native overlay, the repository's React Native overlay evidence runs on iOS simulators and Android emulators, and no macOS lane invokes the command: neither the System Settings replay nor the desktop and recording provider scenarios carry a react-native step.",
802799
),
803800
tvos: tvos.gap('No tvOS-specific React Native inspection command evidence exists yet'),
804801
// R61: no owner fact refuses this command on a browser — its whole device work is one bound
@@ -852,7 +849,7 @@ const COMMAND_COVERAGE_DECLARATIONS = {
852849
),
853850
macos: macos.accepted(
854851
'No command-specific macOS trace evidence exists',
855-
"trace never binds a device runtime: its session-scoped start/stop bookkeeping is platform-neutral and already pinned by the daemon's trace-runtime tests, so a macOS lane would re-run identical code and macOS carries no command-specific trace evidence.",
852+
'trace never binds a device runtime: handleTraceCommand carries no bindDevice or inspectFacts call, its start/stop bookkeeping is pinned by the daemon trace-runtime tests, and no macOS lane invokes the command: neither the System Settings replay nor the desktop and recording provider scenarios carry a trace step.',
856853
),
857854
tvos: tvos.gap('No tvOS-specific trace command evidence exists yet'),
858855
web: web.contract(
@@ -862,7 +859,7 @@ const COMMAND_COVERAGE_DECLARATIONS = {
862859
),
863860
linux: linux.accepted(
864861
'No Linux-specific trace command evidence exists',
865-
"trace never binds a device runtime: its session-scoped start/stop bookkeeping is platform-neutral and already pinned by the daemon's trace-runtime tests, so a Linux lane would re-run identical code and Linux carries no command-specific trace evidence.",
862+
'trace never binds a device runtime: handleTraceCommand carries no bindDevice or inspectFacts call and its start/stop bookkeeping is pinned by the daemon trace-runtime tests, and no Linux lane invokes the command: neither the replay script nor the command-evidence lane carries a trace step.',
866863
),
867864
},
868865
[C.find]: {

‎test/integration/support/coverage-classification.ts‎

Lines changed: 3 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -36,7 +36,9 @@ type CoverageGapRow = {
3636
* - a `known-gap` row is owed work and carries a positive `trackingIssue` naming an open
3737
* issue that owns it, an umbrella or a child of one;
3838
* - an `accepted-gap` row is a permanent limitation: it carries a non-empty `reason` saying
39-
* why no evidence lane will cover it, and no tracking issue.
39+
* why no evidence lane will cover it, and no tracking issue. An accepted-gap reason may
40+
* only claim an absence that holds against the repository's lanes — if a lane runs or
41+
* asserts the command, the row is evidence-citable or owed, never accepted.
4042
*
4143
* This mirrors the `gap:` waiver rule ADR 0011 enforces for interaction guarantees.
4244
*/

0 commit comments

Comments
 (0)