Give collection one publication transaction #96

New Issue

2026-09-27T20:01:02Z

xavierk commented

2026-09-27 20:01:02 +00:00

Parent

Deepen observation-history publication, local-day evidence, and activity selection.

What to build

Make a successful collection run publish its sample, controller-segment context and derived UTC/local evidence as one consistent reader-visible observation history. Give the collection path transaction ownership before adding pending-publication recovery; remove helper-owned intermediate commits without changing the existing acquisition or monitoring controls. This is prefactoring of the successful path, not a claim that failed derivation already has the final recovery behavior.

Acceptance criteria

Controlled acquisition and clock through the normal collection path produce a real temporary observation store whose ordinary TUI/CLI readers see a coherent successful publication.
The collector owns transaction commits; derivation helpers cannot commit partial public changes independently. Repeated successful collection does not double-count measured read/write volume.
Controller-segment and monitoring-period rules, UTC projection inputs, existing store migration and read-only snapshots retain their established behavior.
Add a focused fault-injection test around a publication step proving that no new partial public derived state escapes a failed transaction; do not mislabel this as the complete pending-recovery behavior of ticket 2.
Targeted existing collector, store and reader tests pass in an isolated development environment; no live device, observation store or installed unit is touched.

Blocked by

None (can start immediately).

## Parent [Deepen observation-history publication, local-day evidence, and activity selection](https://git.bongbetic.com/xavierk/Fenris/issues/95). ## What to build Make a successful collection run publish its sample, controller-segment context and derived UTC/local evidence as one consistent reader-visible observation history. Give the collection path transaction ownership before adding pending-publication recovery; remove helper-owned intermediate commits without changing the existing acquisition or monitoring controls. This is prefactoring of the successful path, not a claim that failed derivation already has the final recovery behavior. ## Acceptance criteria - [ ] Controlled acquisition and clock through the normal collection path produce a real temporary observation store whose ordinary TUI/CLI readers see a coherent successful publication. - [ ] The collector owns transaction commits; derivation helpers cannot commit partial public changes independently. Repeated successful collection does not double-count measured read/write volume. - [ ] Controller-segment and monitoring-period rules, UTC projection inputs, existing store migration and read-only snapshots retain their established behavior. - [ ] Add a focused fault-injection test around a publication step proving that no **new partial public derived state** escapes a failed transaction; do not mislabel this as the complete pending-recovery behavior of ticket 2. - [ ] Targeted existing collector, store and reader tests pass in an isolated development environment; no live device, observation store or installed unit is touched. ## Blocked by None (can start immediately).

xavierk added the ready-for-agent label 2026-09-27 20:01:02 +00:00

xavierk referenced this issue

2026-09-27 20:01:04 +00:00

Retain and recover pending publication #97

xavierk added a new dependency 2026-09-27 20:01:05 +00:00

#97 Retain and recover pending publication

xavierk closed this issue

2026-09-27 21:03:40 +00:00

xavierk referenced this issue

2026-10-02 08:22:02 +00:00

Deepen observation-history publication, local-day evidence, and activity selection #95

Sign in to join this conversation.