Current behavior
_globals.gindLin -- the array behind op->indLin -- is allocated fresh on
every solve setup from mv$indLin[[4]] (src/rxData.cpp:6143-6147), but
src/rxSerialize.cpp writes only the scalar indLinN (:142). The array itself
is never round-tripped.
Harmless while codes 3/4 are unreachable. Once the iteration is restored, a
save/restore cycle would come back with an empty convergence set.
Change
Append an int blob at the end of the existing sequence, following the
sWriteIntBlob pattern already used in that file for cov0, splitBolus,
par_sample and linCmtScale. Field order is fixed and guarded by a format
version and a sizeof(rx_solving_options) check -- do not insert mid-list.
Test
Solve, serialize, deserialize, solve again for an indLin()-forcing model;
assert identical output.
Blocked by #1185.
Current behavior
_globals.gindLin-- the array behindop->indLin-- is allocated fresh onevery solve setup from
mv$indLin[[4]](src/rxData.cpp:6143-6147), butsrc/rxSerialize.cppwrites only the scalarindLinN(:142). The array itselfis never round-tripped.
Harmless while codes 3/4 are unreachable. Once the iteration is restored, a
save/restore cycle would come back with an empty convergence set.
Change
Append an int blob at the end of the existing sequence, following the
sWriteIntBlobpattern already used in that file forcov0,splitBolus,par_sampleandlinCmtScale. Field order is fixed and guarded by a formatversion and a
sizeof(rx_solving_options)check -- do not insert mid-list.Test
Solve, serialize, deserialize, solve again for an
indLin()-forcing model;assert identical output.
Blocked by #1185.