@@ -127,32 +127,30 @@ describe("Loading `on` beside a keyed Show around the boundary; a boundary is ne
127127 // first landing, mid-flight, after the second landing]. Static and callback
128128 // children are re-evaluated by `Show` at the key change: one more mount,
129129 // mid-flight. `on` never remounts.
130- // Over `latest(count)` that is two mounts mid-flight (design ruling R-5,
131- // #4365): `latest()` opens no transaction for its verdict, so the
132- // condition's first pass is the frame's — `Show` mounts A for key 2 in the
133- // frame, and the frame parks on data, holding that mount. The parked flush
134- // re-runs the `latest()` reader once in the verdict lane, and `Show` mounts
135- // A again there: the lane's mount is the one shown (the early fallback),
136- // and the held frame's is discarded at the lane's reveal. (Before R-5 the
137- // read opened the flush's transaction and seated the pass in the lane at
138- // once: one mount.) Why 3: the parked flush re-derives the pass that read
139- // `latest(count)` once, as a verdict reader (D-2, A10), and a keyed `Show`
140- // whose condition is that read mounts its branch on that re-derive. That is
141- // the cost of keying a mount on a `latest()` read at a park, not a bug
142- // (maintainer ruling on #4365).
130+ // Over `latest(count)` that one mount is mid-flight: `latest()` opens no
131+ // transaction for its verdict (R-5, #4365), so the condition's first pass
132+ // is the frame's — `Show` mounts A for key 2 in the frame, and the frame
133+ // parks on data. The parked flush re-derives the pass that read
134+ // `latest(count)` once, as a verdict reader (D-2, A10); `latest()` serves
135+ // the staging on both runs, so the key is the same and the one child stays
136+ // mounted. Today the re-derive voids the provisional owner and compares
137+ // against the last COMMITTED `when`, so the same key mounts again
138+ // ([1, 3, 3]) — the known bug #4417, pinned as `test.fails` at its target.
143139 // (A zero-arg function child is an accessor — evaluated once, not remounted
144140 // per key — and is deliberately not pinned here.)
145141 const matrix : Array < [ Variant , string , [ number , number , number ] , string [ ] ] > = [
146- [ "show-static-latest" , EARLY , [ 1 , 3 , 3 ] , [ ] ] ,
147- [ "show-callback-latest" , EARLY , [ 1 , 3 , 3 ] , [ ] ] ,
142+ [ "show-static-latest" , EARLY , [ 1 , 2 , 2 ] , [ ] ] ,
143+ [ "show-callback-latest" , EARLY , [ 1 , 2 , 2 ] , [ ] ] ,
148144 [ "show-static-committed" , HELD , [ 1 , 2 , 2 ] , [ ] ] ,
149145 [ "show-callback-committed" , HELD , [ 1 , 2 , 2 ] , [ ] ] ,
150146 [ "on-committed" , HELD , [ 0 , 0 , 0 ] , [ "LOADING_ON_OUTSIDE_HOLD" ] ] ,
151147 [ "on-latest" , EARLY , [ 0 , 0 , 0 ] , [ ] ]
152148 ] ;
153149 for ( const [ variant , midFlight , expectedMounts , expectedCodes ] of matrix ) {
154150 const remounted = expectedMounts [ 1 ] > expectedMounts [ 0 ] ;
155- test ( `${ variant } : mid-flight ${ midFlight === EARLY ? "reveals fallback early" : "held" } , A ${
151+ // Known bug #4417: the park re-derive voids the provisional owner, so the same key mounts again ([1,3,3]). Target [1,2,2] via in-place re-derive (reader-seat S1).
152+ const it = variant . startsWith ( "show-" ) && variant . endsWith ( "-latest" ) ? test . fails : test ;
153+ it ( `${ variant } : mid-flight ${ midFlight === EARLY ? "reveals fallback early" : "held" } , A ${
156154 remounted ? "remounted" : "not remounted"
157155 } at the key change`, async ( ) => {
158156 const { div, setCount, dispose, mounts } = setup ( variant ) ;
0 commit comments