Reproduction
With deterministic=.true., initialize the sampler twice with different explicit ran_seed values and compare the generated random streams.
Expected
Equal explicit seeds reproduce the stream. Different explicit seeds produce different streams. Omitting ran_seed preserves the historical deterministic seed 42.
Actual
The deterministic reset overwrites the configured seed, so distinct ran_seed values generate the same stream.
Environment
SIMPLE main before PR #452; GNU Fortran deterministic build.
Evidence
The behavioral regression in PR #452 fails with different ran_seed values produce the same random stream before the fix and passes after the reset uses the configured seed.
Reproduction
With
deterministic=.true., initialize the sampler twice with different explicitran_seedvalues and compare the generated random streams.Expected
Equal explicit seeds reproduce the stream. Different explicit seeds produce different streams. Omitting
ran_seedpreserves the historical deterministic seed 42.Actual
The deterministic reset overwrites the configured seed, so distinct
ran_seedvalues generate the same stream.Environment
SIMPLE
mainbefore PR #452; GNU Fortran deterministic build.Evidence
The behavioral regression in PR #452 fails with
different ran_seed values produce the same random streambefore the fix and passes after the reset uses the configured seed.