pj 3caf53cfb6 test(chrome): one launch preamble, and parity now pins ground truth ...

48 of 51 call sites shared one eight-line block that opened a fixture server,
launched chrome and checked the error. The three left are the launch tests
themselves, where what the helper hardcodes is the subject.

The audit called four tests superseded by fact_parity_test.go. They were not:
parity asserts the two producers agree and that each fact appears both ways,
never that either reading is right, so deleting them would have removed the only
thing anchoring the pair to reality. The values they pinned are now asserted
inside parity against both producers, and the shadow fixture has ground truth
where it had none.

2026-08-22 21:31:13 +05:30

..

testdata

record what the model picker did, and make both policies see the same actions (#74)

2026-08-19 11:07:49 +05:30

driver_test.go

test(chrome): one launch preamble, and parity now pins ground truth

2026-08-22 21:31:13 +05:30

driver.go

fix(chrome): a failed metrics round trip is an error, not zero bytes

2026-08-22 21:07:15 +05:30

element_state_test.go

test(chrome): one launch preamble, and parity now pins ground truth

2026-08-22 21:31:13 +05:30

fact_parity_test.go

test(chrome): one launch preamble, and parity now pins ground truth

2026-08-22 21:31:13 +05:30

launch_test.go

fix ltl semantics and unify action enumeration (#71)

2026-08-12 18:06:04 +05:30

logs_test.go

correctness fixes from the first real folio dispatch, and spec authoring skills (#82)

2026-08-16 13:23:53 +05:30

picker_stream_test.go

test(chrome): one launch preamble, and parity now pins ground truth

2026-08-22 21:31:13 +05:30

selector_parity_test.go

test(chrome): one launch preamble, and parity now pins ground truth

2026-08-22 21:31:13 +05:30

translate_test.go

record what the model picker did, and make both policies see the same actions (#74)

2026-08-19 11:07:49 +05:30

translate.go

docs: stop labelling current behaviour as legacy

2026-08-22 21:11:13 +05:30