v0.2.54.0 feat: a run is refused up front when saved studies and test replays fill the limit, and a server replay has storage to land in (Cycle 0059 pt1) by ryota-murakami · Pull Request #148 · laststance/codebook

test: a history saved by v0.2.53.0 is a committed fixture that must list and open unchanged

ryota-murakami

committed

24c3cc8

feat: one storage descriptor per history profile, a node-server profile with its writer rules and bounds, and a writer that drains accepted entries before its seal

ryota-murakami

committed

b5d5670

feat: one saved-study limit counts saved studies, test histories and server histories together, in both directions

ryota-murakami

committed

0b5db5f

test: journal integrity, structured seal, redaction at rest and profile separation of the server history store

ryota-murakami

committed

fba177d

chore: keep the storage foundation inside the dead-code and duplication baselines (moved writer identity, shared pair masker, pt3-only members suppressed with a reason)

ryota-murakami

committed

160600a

chore: record the new history test files' timings so the shard plan splits the 20-second bounds file instead of weighing it as one second

ryota-murakami

committed

7fe6a6b

docs: the saved study limit is shared in both directions, so a new study's save counts the test histories too

ryota-murakami

committed

6336396

fix: a server seal closes the writer to new appends, writes end once and never publishes a seal without end

ryota-murakami

committed

5feeab0

fix: a stop lists at most one unreadable value per inspection, and a stop whose executed source was not stored seals the capture incomplete

ryota-murakami

committed

5addfe6

fix: a history directory that cannot be read counts no history for a browser save, as an unreadable Node history store never blocked one

ryota-murakami

committed

c73b44c

chore: drop the stale seal suppression and the dead-code baseline ids that no longer match a finding

ryota-murakami

committed

3e4d15e

fix: a refused append with a long cause is cut to the seal's bound, so the partial server history seal is always written

ryota-murakami

committed

1d8e31f

fix: a run's start check and a recording's start count the test and server histories that share the saved-study limit

ryota-murakami

committed

1c3c3d9

test: the shared journal rules refuse a withdrawal after the bytes left, a request after the terminal entry and output after its truncation, in writer, scan and read

ryota-murakami

committed

dd75f1e

docs: the root README names 0.2.54.0 and the three parts of Cycle 0059

ryota-murakami

committed

1d79c79

chore: bump version and changelog (v0.2.54.0)

ryota-murakami

and

claude

committed

1972b17

Navigation Menu

Repository navigation

v0.2.54.0 feat: a run is refused up front when saved studies and test replays fill the limit, and a server replay has storage to land in (Cycle 0059 pt1) - #148

v0.2.54.0 feat: a run is refused up front when saved studies and test replays fill the limit, and a server replay has storage to land in (Cycle 0059 pt1)#148
ryota-murakami merged 16 commits into
mainfrom
codex/cycle-0059-pt1

Commits on Oct 28, 2026

test: a history saved by v0.2.53.0 is a committed fixture that must list and open unchanged

feat: one storage descriptor per history profile, a node-server profile with its writer rules and bounds, and a writer that drains accepted entries before its seal

feat: one saved-study limit counts saved studies, test histories and server histories together, in both directions

test: journal integrity, structured seal, redaction at rest and profile separation of the server history store

chore: keep the storage foundation inside the dead-code and duplication baselines (moved writer identity, shared pair masker, pt3-only members suppressed with a reason)

chore: record the new history test files' timings so the shard plan splits the 20-second bounds file instead of weighing it as one second

docs: the saved study limit is shared in both directions, so a new study's save counts the test histories too

fix: a server seal closes the writer to new appends, writes end once and never publishes a seal without end

fix: a stop lists at most one unreadable value per inspection, and a stop whose executed source was not stored seals the capture incomplete

fix: a history directory that cannot be read counts no history for a browser save, as an unreadable Node history store never blocked one

chore: drop the stale seal suppression and the dead-code baseline ids that no longer match a finding

fix: a refused append with a long cause is cut to the seal's bound, so the partial server history seal is always written

fix: a run's start check and a recording's start count the test and server histories that share the saved-study limit

test: the shared journal rules refuse a withdrawal after the bytes left, a request after the terminal entry and output after its truncation, in writer, scan and read

docs: the root README names 0.2.54.0 and the three parts of Cycle 0059

chore: bump version and changelog (v0.2.54.0)