Elamur Groupantswasi-clock-poll
The groupant beside wasi-host/src/lib.rs in the Mova repository: what the code computes, what it never does, how to go back from its outputs, its invariants, cost and checks.
Facts
Implementation material. A Mova code model describes this code at one source revision: partial models, finite checks and their stated scope. It is not a reviewed general exposition, and its labels are shown as written, never raised.
Components
| component | label | where |
|---|---|---|
| approximant | partial | approximant |
| ast | partial | ../syntax.json |
| cost | symbolic source derived | invariant |
| descriptant | present | descriptant |
| invariant | partial | invariant |
| tests | executed bounded checks | tests |
Obernants
The typed relations around the forward model: the behaviour complement (what the code never does), the reverse reconstruction (every input compatible with an output) and specialised views.
| id | type | label | where |
|---|---|---|---|
behavior_complement | behavior_complement | partial source derived model | behavior_complement |
reverse_reconstruction | reverse_reconstruction | partial source derived model | reverse_reconstruction |
clock_coordinates | clock_coordinates | partial source derived model | clock_coordinates |
timer_ties | timer_ties | partial source derived model | timer_ties |
random_zero | random_zero | partial source derived model | random_zero |
partial_output | partial_output | partial source derived model | partial_output |
Model text
descriptant
wasi-clock-poll
String sizes write count then total nul-terminated bytes; strings_get writes pointer then bytes then terminator. clock validates id<=3, reads now and advances before put. random validates range before consuming ceil(len/8) splitmix words. poll validates1..4096 subscriptions, prioritizes fd events; otherwise advances minimum timer and emits every minimum tie.
Epoch 2026-10-31
Prospective review recorded before edits. Source review found late output FAULT effects and weaker-than-capability rights metadata; these are modeled as current behavior rather than silently repaired. Only bounded new test blocks were added; production code remains unchanged. Actual finite results and exact restoration receipts live under $MOVA_DATA/runs/octopus-wave31-wasi-2026-10-31. No full mlab guest, frozen suite, GPU or corpus run. This node retains its scoped observations for future refinement.
Independent review narrowed reverse reconstruction: emitted event bytes carry userdata/type, not timestamps. The minimum timer duration is recoverable only with before/after clock observations, timer-only execution and nonsaturating advancement. Event bytes alone leave durations ambiguous.
approximant
Forward relation
String sizes write count then total nul-terminated bytes; strings_get writes pointer then bytes then terminator. clock validates id<=3, reads now and advances before put. random validates range before consuming ceil(len/8) splitmix words. poll validates1..4096 subscriptions, prioritizes fd events; otherwise advances minimum timer and emits every minimum tie.
Scope: wasi-clock-poll. State and transitions follow the authoritative source, including partial effects before errors. No executable approximate replacement is attached: categorical errno/effects and module execution lack a useful validated percentage metric; future numerical profiles may only use1/2/5/10% caps after domain/reference evidence.
invariant
Invariant and cost
String sizes write count then total nul-terminated bytes; strings_get writes pointer then bytes then terminator. clock validates id<=3, reads now and advances before put. random validates range before consuming ceil(len/8) splitmix words. poll validates1..4096 subscriptions, prioritizes fd events; otherwise advances minimum timer and emits every minimum tie.
The invariant is conditional on the admitted transition: A valid timer-only poll cannot omit one of the minimum ties when output storage fits; fd-ready poll cannot advance timer clock. Invalid random range cannot consume RNG state.
Cost: O(argument bytes+count), O(random len) allocated len+8; poll O(nsubs) space/time plus O(events) writes.
No wall-time gain, universal resource bound or atomic rollback is inferred. External Vfs/wasmi/native-mlab work remains separately charged.
models
Typed model catalog
behavior_complement
Subject: wasi-host/src/lib.rs at SHA-256 91213ce9038107912c2dd707c3a6639a2d723cdfceb02981a166a6853800248e, scope wasi-clock-poll. Domain: the source transition and bounded fixtures described in approximant.md; readable valid host resources where required. Observation: returned values, byte coordinates, state/effect order, not total-system semantics. Operator/relation: A valid timer-only poll cannot omit one of the minimum ties when output storage fits; fd-ready poll cannot advance timer clock. Invalid random range cannot consume RNG state.
Status: partial source-derived relation; complement is a scoped exclusion hypothesis, reverse preserves all compatible antecedents described rather than a chosen inverse. Assumptions: production branch order, declared coordinate/resource premises; no completeness claim for untested inputs. Evidence: tests.md distinguishes actual execution from source review; finite observations can falsify but not prove the relation.
reverse_reconstruction
Subject: wasi-host/src/lib.rs at SHA-256 91213ce9038107912c2dd707c3a6639a2d723cdfceb02981a166a6853800248e, scope wasi-clock-poll. Domain: the source transition and bounded fixtures described in approximant.md; readable valid host resources where required. Observation: returned values, byte coordinates, state/effect order, not total-system semantics. Operator/relation: Event bytes recover emitted userdata and event type, but contain no timer timestamp or duration. For timer-only execution, observing both before and after virtual clock values recovers the minimum relative duration only when clock advancement does not saturate; larger timer durations remain ambiguous. Descriptor-ready events do not establish any timer duration. Random output truncates the final64-bit word, hiding remaining bytes.
Status: partial source-derived relation; complement is a scoped exclusion hypothesis, reverse preserves all compatible antecedents described rather than a chosen inverse. Assumptions: production branch order, declared coordinate/resource premises; no completeness claim for untested inputs. Evidence: tests.md distinguishes actual execution from source review; finite observations can falsify but not prove the relation.
clock_coordinates
Subject: wasi-host/src/lib.rs at SHA-256 91213ce9038107912c2dd707c3a6639a2d723cdfceb02981a166a6853800248e, scope wasi-clock-poll. Domain: the source transition and bounded fixtures described in approximant.md; readable valid host resources where required. Observation: returned values, byte coordinates, state/effect order, not total-system semantics. Operator/relation: id0 returns epoch time;1..3 return now-EPOCH_NS. Invalid id returns INVAL without advance; valid id with bad out advances then FAULT.
Status: partial source-derived relation; complement is a scoped exclusion hypothesis, reverse preserves all compatible antecedents described rather than a chosen inverse. Assumptions: production branch order, declared coordinate/resource premises; no completeness claim for untested inputs. Evidence: tests.md distinguishes actual execution from source review; finite observations can falsify but not prove the relation.
timer_ties
Subject: wasi-host/src/lib.rs at SHA-256 91213ce9038107912c2dd707c3a6639a2d723cdfceb02981a166a6853800248e, scope wasi-clock-poll. Domain: the source transition and bounded fixtures described in approximant.md; readable valid host resources where required. Observation: returned values, byte coordinates, state/effect order, not total-system semantics. Operator/relation: Equal minimum durations all emit events in input order; any type1/2 subscription suppresses timer advancement.
Status: partial source-derived relation; complement is a scoped exclusion hypothesis, reverse preserves all compatible antecedents described rather than a chosen inverse. Assumptions: production branch order, declared coordinate/resource premises; no completeness claim for untested inputs. Evidence: tests.md distinguishes actual execution from source review; finite observations can falsify but not prove the relation.
random_zero
Subject: wasi-host/src/lib.rs at SHA-256 91213ce9038107912c2dd707c3a6639a2d723cdfceb02981a166a6853800248e, scope wasi-clock-poll. Domain: the source transition and bounded fixtures described in approximant.md; readable valid host resources where required. Observation: returned values, byte coordinates, state/effect order, not total-system semantics. Operator/relation: len0 consumes no generator state, provided zero-length range valid; invalid range consumes no state.
Status: partial source-derived relation; complement is a scoped exclusion hypothesis, reverse preserves all compatible antecedents described rather than a chosen inverse. Assumptions: production branch order, declared coordinate/resource premises; no completeness claim for untested inputs. Evidence: tests.md distinguishes actual execution from source review; finite observations can falsify but not prove the relation.
partial_output
Subject: wasi-host/src/lib.rs at SHA-256 91213ce9038107912c2dd707c3a6639a2d723cdfceb02981a166a6853800248e, scope wasi-clock-poll. Domain: the source transition and bounded fixtures described in approximant.md; readable valid host resources where required. Observation: returned values, byte coordinates, state/effect order, not total-system semantics. Operator/relation: String pointer cells and prior events can be written before a later memory fault; no all-or-none output promise.
Status: partial source-derived relation; complement is a scoped exclusion hypothesis, reverse preserves all compatible antecedents described rather than a chosen inverse. Assumptions: production branch order, declared coordinate/resource premises; no completeness claim for untested inputs. Evidence: tests.md distinguishes actual execution from source review; finite observations can falsify but not prove the relation.
tests
Checks and limits
Named actual-source tests run twice per debug/release; exact commands and numeric statuses retained externally. Test success is finite evidence only.
Exact names: tests::clock_is_virtual_and_random_is_seeded, tests::wave31_memory_poll_and_descriptor_boundaries, tests::wave31_late_faults_keep_prior_effects.
Memory tests bypass linker ABI; tiny empty/unreachable guests exercise engine outcomes only. The CLI test uses isolated import/export/journal fixtures.
Five compiled production/helper-source controls: append offset, virtual clock advance, trap status, CLI export gate and newest-mtime max. Each compiles, fails its selected assertion, restores exact original bytes and passes again. These controls cover their target relations, not every node. No assertion-only mutants. See external source-test-map.json, native-commands.json, verify.rs and numeric/raw receipts in $MOVA_DATA/runs/octopus-wave31-wasi-2026-10-31. The example main and frozen mlab paths are never executed.
This page in Markdown → https://md.groupants.elamur.ai/free/en/mova/wasi-host/src/lib.rs.groupant/wasi-clock-poll.groupant/