v0.1.7

Changelog

• ad9a0ad Merge pull request #101 from curiouspub/the-package-moves-to-0.1.7
• 998ae58 release: the package moves to 0.1.7, and the README with it
• 6cf0f26 Merge pull request #99 from curiouspub/the-symlink-warning-asks-about-links-the-build-reads
• 8e52895 pack: a rooted link target is outside the project on every platform
• 023c4d4 Merge branch 'main' into the-symlink-warning-asks-about-links-the-build-reads
• 68a018d Merge pull request #98 from curiouspub/the-name-check-reads-the-public-folder
• 949f75c Merge pull request #100 from curiouspub/the-terminal-failure-shows-its-deploy-id
• e2e0374 ui: a failure shows its deploy id in the terminal, under the failure id
• 7e68ae1 pack: a symlink warns only where the build reads, naming where it points
• c1a5ca5 pack: the page check itself, which the commit before this one describes
• 4dc7fed pack: a page's name is checked too, without its bracketed parameters
• f934914 pack: the name check reads only the public folder, where names survive the build

v0.1.6

Changelog

• cd5bde8 Merge pull request #97 from curiouspub/the-package-moves-to-0.1.6
• b0cbc74 release: the package moves to 0.1.6, and the README with it
• bcc2297 Merge pull request #95 from curiouspub/the-copy-names-its-channel-and-its-limits
• 99547c5 Merge branch 'main' into the-copy-names-its-channel-and-its-limits
• e3bfa06 Merge pull request #96 from curiouspub/the-harness-numbers-connections-on-arrival
• 7a18002 test(flow): the events harness numbers its connections on arrival
• d2e00ac copy: the internal-fault message names the support address, and the READMEs state the output limit and the first-run fetch
• 9b74ff9 Merge pull request #94 from curiouspub/a-check-is-read-from-a-file
• cc09083 Merge pull request #93 from curiouspub/the-stale-sentences-corrected
• 612bccc docs: a check's output is read from a file, never through a pipe
• bb55ff7 wire: the status and phase comments say which values are produced today
• acb63f0 Merge pull request #92 from curiouspub/the-release-names-a-failed-signer-install
• 63591e9 release: the plan prints the recovery from a failed signer install
• f094337 Merge pull request #91 from curiouspub/the-read-side-windows-re-measured
• fcc0566 test(timing): the read-side windows, re-measured under the current arming
• 8741ed5 Merge pull request #90 from curiouspub/the-probe-prints-every-record
• 28da863 test(flow): every probe pass prints its whole record, ready to paste
• 56571bc Merge pull request #89 from curiouspub/the-probe-measures-from-the-response
• aac6f6d test(flow): the stream probe starts its clock where the watchdog now arms
• 0ffa51b Merge pull request #88 from curiouspub/the-stream-has-one-bound-per-phase
• 18cc8f0 fix(stream): one bound per phase, and the stall window starts on the response
• deffaea Merge pull request #87 from curiouspub/the-stream-reader-can-show-its-timeline
• c4157db test(flow): a red stall row shows both timelines, the server's and the reader's

v0.1.5

Changelog

• d1bc6ab Merge pull request #86 from curiouspub/the-package-moves-to-0.1.5
• 26cfd85 release: the package moves to 0.1.5, and the README with it
• 9b53b7b Merge pull request #85 from curiouspub/the-cask-is-removed
• 522f255 release: the cask is removed, and the tap's copy goes with it
• 22be0f7 Merge pull request #84 from curiouspub/the-notary-log-finds-this-runs-submissions
• 67c5d20 release: the notary log finds this run's submissions, and a timeout is re-run

v0.1.4

Changelog

• 8c23224 Merge pull request #83 from curiouspub/the-package-moves-to-0.1.4
• e4444ee release: the package moves to 0.1.4, and the README with it
• 6f57bbb Merge pull request #82 from curiouspub/the-macos-binaries-are-signed-and-notarised
• dae8aee release: the macOS binaries are signed with a developer certificate and notarised

v0.1.3

Changelog

• 518855d Merge pull request #81 from curiouspub/the-package-moves-to-0.1.3
• 57bce68 release: the package moves to 0.1.3, and the README with it
• d42f05c Merge pull request #80 from curiouspub/the-cask-migration-names-both-commands
• 877a198 Merge branch 'main' into the-cask-migration-names-both-commands
• 8c83201 test(readme): the README names the package's declared version, not the newest tag
• aa87a97 Merge pull request #79 from curiouspub/the-linux-check-follows-the-formula
• bbd4bfa Merge pull request #78 from curiouspub/the-readback-waits-for-the-registry
• 41d711a release: the move off the cask names both commands, and the README names the newest release
• 9c6664f release: the Linux install check follows the formula job, not the whole run
• dd5ee0b release: the readback waits for the registry, and every registry read is live

v0.1.2

Changelog

• 0019b12 Merge pull request #77 from curiouspub/the-version-agrees-before-the-tag
• 1938fc3 release: the version is checked for agreement before the tag, not after

v0.1.1

Where this release was published

v0.1.1 is on GitHub Releases and in the Homebrew tap only. The binaries, checksums.txt and its signature bundle are attached below. The tap's formula (brew install curiouspub/tap/curiouspub) and its cask both install 0.1.1.

The npm package was not published for this version, and npm install -g curiouspub still installs 0.1.0. The package's own version was left at 0.1.0 in the tagged commit. The publish step compares that version with the tag, and it refused, correctly: a package built from this commit would have asked for release files that do not exist. The tag could not be moved once its binaries were public.

v0.1.2 publishes the package. The release script now refuses a version that the commit does not declare, before any tag is created.

Changelog

• f25c877 Merge pull request #76 from curiouspub/the-tap-has-a-formula
• 1a6ad7e release: the Homebrew channel is a formula, written from the release's own checksums
• 073964f Merge pull request #75 from curiouspub/three-codes-for-three-causes
• 0854f98 feat(wire): three codes say whose side a stopped build or a refused publish is on
• 7cf3bec Merge pull request #74 from curiouspub/one-way-to-write-the-address
• cbae668 docs: one support address, written one way
• 41ab8e9 Merge pull request #73 from curiouspub/the-release-chain-holds-its-own-checksums
• 8db50db ci: the release chain holds its own checksums, and every install names its version
• f36ac06 Merge pull request #72 from curiouspub/the-start-says-when
• 4bef061 fix(deploy): a rate-limited start says when, and waits
• b6a144c Merge pull request #71 from curiouspub/the-status-is-one-phrase
• f025b0e docs(readme): the status is one phrase — testing, open, limited
• 373c7ec Merge pull request #70 from curiouspub/the-fetch-moves-to-first-invocation
• b6ccde9 fix(npm): a spawn that throws synchronously is refused, not printed as a trace
• edbf0f6 test(npm): the fetch is asserted everywhere, running the result where it can be
• 4c28ad6 fix(npm): the binary is fetched on first run when the install could not
• f6e0995 Merge pull request #69 from curiouspub/the-deadline-is-a-parameter
• 4a8e1b8 perf(leakscan): the deadline is a parameter, and the shipped number keeps a row
• 1f6bcc0 Merge pull request #68 from curiouspub/the-leak-scan-hangs-in-fork
• 783d061 fix(leakscan): the stub holds its working directory too, not only its image
• 8189db9 fix(leakscan): the blocking stub outlives the test, so its directory is cleaned best-effort
• b35b641 fix(leakscan): a git call that never returns fails in seconds, naming what it read
• ab5c597 Merge pull request #67 from curiouspub/the-cask-hook-uses-the-current-form
• aca3611 fix(release): the cask lints to exactly the deprecation, and the fix triggers itself
• aa74155 Merge pull request #66 from curiouspub/a-failure-carries-its-origin
• c6e43d4 feat(wire),fix(flow,ui): a platform limit is neither side's fault
• 9fa781a feat(wire),fix(flow,ui): a failed build says whose fault it was
• 54d5233 Merge pull request #65 from curiouspub/the-platform-half-is-present-tense
• f830ac6 docs(readme),test(guard): the platform half is present tense, and held to a recording
• 57c19a2 Merge pull request #64 from curiouspub/health-response-joins-the-contract
• cecff94 feat(wire): a liveness response joins the contract, with its all-zero state pinned
• 6e41a95 Merge pull request #63 from curiouspub/the-floor-is-stated-where-it-is-claimed
• 509fdd4 docs(readme): "What this is" is conditional too, for the reason the rest is
• 69848a4 test(flow): each row states the floor it claims, and a row holds them to it
• 3a31014 Merge pull request #62 from curiouspub/the-note-has-one-home
• 0645117 docs(timing): the carried entry points at its home instead of restating it
• 0d253d0 Merge pull request #61 from curiouspub/the-page-says-what-is-published
• 338bf7f docs(readme): the page says what is released and what is not running
• ef06d04 test(guard): a released tree does not say it is unreleased

v0.1.0

Changelog

• 38ad79c Merge pull request #60 from curiouspub/the-snapshot-signs-before-the-tag
• d90d995 ci(release): the snapshot signs, so the release's arguments are exercised before a tag
• 205594b Merge pull request #59 from curiouspub/the-publish-exchanges-its-identity
• f8e6e32 ci(release): the wrapper publish exchanges its identity, and its package manager is ensured
• bfa6619 Merge pull request #57 from curiouspub/the-guard-names-the-release
• 16e0631 test(readme-contract): only a final version retires the disclaimer
• f86e0f4 Makefile: release-log captures a release run
• baca1dd test(readme-contract): the publication row names the release it waits for
• 65b21f9 Merge pull request #58 from curiouspub/the-buffer-is-sampled
• 567e4f5 test(socketpin): the mutation record says what the run printed
• ac2ee8a test(socketpin): the buffer is sampled while the body moves
• 80364a1 Merge pull request #56 from curiouspub/the-exit-test-runs-itself
• 8e85c7e test(exit-test): ask the question where it can be answered, and name the fix in the refusal
• c90e724 build,test: the exit test runs itself
• 05a8c3e test(fixtures): a project whose srcDir is computed, and no pages to find
• fcc1ee2 Merge pull request #55 from curiouspub/the-readme-quotes-the-catalog
• 286ebb7 docs: a push is chained to the surface check, and the run that taught us
• e7a5d0c test(guard),build: read the wrapper's copy with a parser, and let the module name itself
• f3b148d test(leakscan,guard),build: a rule may be added; retiring one needs a record
• d6a9928 docs(README),test(flow): a warning is not a refusal, and a row says so
• 79bffbc test(guard): a one-directional rule passes rather than skips
• 217bae3 test(guard): let the copy audit reach the copy it was not reading
• 10f3ebe docs(README),test(guard): hold three of the README's claims to the tree
• 832b855 test(guard,pack): make five rows check what their names claim
• ab65bac docs(README),test(guard): say what a composed message names
• e326d2b docs(guard): state the conclusion without pointing at where it came from
• 3ce8a58 docs: correct the exit-codes note — cancelling also exits 0
• 6bde5c9 docs: rewrite the README for a first-timer, quoting the catalog throughout
• e0bba49 docs: move the spelling-per-file rule into the public CLAUDE.md
• 1d9e7fd test(guard,pack): hold the README's claims to their own sources
• 6c6b4bc Merge pull request #54 from curiouspub/the-stage-is-declared
• 83eaa1e Document every declared stage in one sentence, and publish the twelve in the catalog's note
• 1f2f706 test(guard): a site that declares no stage is told so, not that it declared an empty one
• ea653aa test(guard): a composed headline is held by how it is built: one constructor and one shape per id and stage
• d07ea54 test(guard): the headline rows: every site declares a stage, one What per id and stage, and the README quotes only what the catalog carries
• 03fa96c test(guard),build: the catalog records the stages sites declare and quotes the What each one writes
• 51095fd feat(ui),feat(flow),test(guard): a failure declares the stage a person was in, at every site, the way it declares its id
• 7d92603 Merge pull request #50 from curiouspub/the-blob-universe
• 2c1c977 merge: bring the branch up to main again, which now carries the tokeniser-cost row's provisional ceiling
• 5cea657 Merge pull request #53 from curiouspub/the-tokeniser-publishes-its-ratio
• b397fed test(citations): the tokeniser-cost row measures in milliseconds, publishes its ratio every run, and its ceiling is marked provisional
• d495f3d merge: bring the branch up to main, which now carries the re-ruled quiet-stream record and the queue's single run
• e9fa65e Merge pull request #52 from curiouspub/the-queue-runs-once
• 8712a42 Merge pull request #51 from curiouspub/the-quiet-stream-retake
• 96989d5 test(timing): the quiet-stream darwin record is re-ruled from every hosted reading, and its budget is held as a ceiling
• 8949d66 ci: the merge queue's own branches no longer trigger a second run of the whole matrix
• 249d4d7 test(leakscan): two claims that were true and proved by nothing now each have a row
• 6b914c9 test(leakscan): a fixture's value is git's stdout, not its stderr too
• 82325d5 docs: the history scan's cost figure keeps one home, beside its target
• fbeb04a merge: bring the branch up to main, which carries the merge-queue flake fix it was missing
• be60756 fix(leakscan): membership is objects and not commits, and a NUL stops hiding the rest of a blob
• 9d3703d fix(leakscan),fix(rules): a blob is checked at every path it was published at, and a manifest that cannot be read stops being clean
• 72d1512 feat(leakscan): the history is scanned as blobs reachable from origin, and a ledger records what is already out
• ad836d5 refactor(leakcheck): one engine answers "is this a leak", and it takes the path as an argument
• 6c21b49 Merge pull request #49 from curiouspub/the-windows-retake
• 97cdb9d feat(rules): every rule in the four manifests carries a stable id, and a report names it instead of the rule
• 1361e3f fix(test): the range fixture's clone disables background gc, which was ejecting pull requests from the merge queue
• 68aa9c2 test(timing),docs: the windows keep-alive record is retaken as the worse of two, and two things that cannot be read back from inside the repo
• 1b9c0e4 Merge pull request #48 from curiouspub/the-failure-catalog
• 03ff16e fix(guard): a site key is a property of the repository, not of the host — and a ledger entry that matches nothing must say so
• f03123e test(guard): every site carrying an id carries the same action, and the census cannot see it
• f0974eb feat(ui),feat(flow),test(guard),build: the four rulings — a family split, an obligation ledger, and a catalog anybody can read
• 5f5d47e fix(guard),test(guard),test(flow),refactor(flow): ten more, and two of them were fc4c811's own regressions
• fc4c811 fix(guard),test(guard),test(flow): the checker reads types rather than spellings, and six more holes a cold review found
• 71b93c4 test(guard): the census row reads the checker's count without inheriting its verdict
• 94754ed fix(ui),test(guard): four failures had no id, and the checker that certified them visited calls only
• 6c1b33a test(flow): the preflight goldens carry the aggregate's id, and the filter that hid them
• e6834c1 feat(ui),feat(mcp),test(guard): the id reaches both readers, and the contract is checked rather than asserted
• 1c7ffc8 fix(flow): the three adjudicated action corrections, each with a row at its caller
• c26f0d7 feat(ui),feat(check),feat(flow): every failure carries a stable id, and joining the waitlist stops being one
• 665679f Merge pull request #47 from curiouspub/the-gate-checks-itself
• 3971827 build,test(guard): the gate names what it skips, and refuses to run on the trunk
• 893146e Merge pull request #46 from curiouspub/the-four-tools
• e3b4d88 fix(ui),fix(flow),test(guard): a setting named in a failure is named in its Why
• 8c5d24b feat(ui),feat(flow),feat(mcp): the next step travels as a value, not as a sentence about a terminal
• 05ea05f feat(ui),feat(flow),feat(mcp): a refused deploy keeps the id its follow-up call needs
• 2c9cdff test(mcp),test(flow): the status a build has never heard of, and a mutation note that said eight
• 30f74f7 test(mcp): the capacity fixture matches by tail like its neighbours
• 834d996...