docs: wizard and tree decisions, audit closures and open nits, session state; four releases log the sign-in address, not eleven #54

2026-11-11T08:46:06Z

james.maes commented

2026-11-11 08:46:06 +00:00

Documentation only: the decisions taken while planning the connection wizard and server tree, audit closures and open review nits from pull requests 44 to 53, the session state, and what waits for the owner. It also corrects a merged statement about which releases log the sign-in address.

Correction of a merged statement

Pull request 53's changelog entry, its description and audit row U22 said every release from 0.2.0-rc.1 to 1.2.1 writes the sign-in address to the companion's log when Windows cannot start the browser. By the code at the tags it is four releases: 1.1.0, 1.1.1, 1.2.0 and 1.2.1 (and snapshot-2.0.0-rc.1). The logging came with aaa5051, which only those tags contain.
The seven releases from 0.2.0-rc.1 to 1.0.1 start the browser the same way but catch the failure and show a fixed text.
Corrected here: the changelog entry, U22, docs/TODO.md, docs/SESSION-STATE.md. Corrected outside this change: the snapshot's release notes on Gitea, and a comment on pull request 53.

ADR-0005: "Resolved while planning the build, 2026-11-10/11"

Decisions G1 to G21, each marked built, built in part or not built, with the code read to say so.
The sign-in provider rule is recorded as the merged contract enforces it: servers.add and trust.set state the provider the operator was shown; a missing value means the check named none; any disagreement is refused as provider_changed.
G7 (which control is the tree) and G11 (re-sending a credential after a refused bind) keep their conditions; neither is met yet.
G19: driving the layout host through UI Automation, and adding the layout tests to the solution when the interactive suites are built. The plan called the second half the owner's call; it is recorded as taken under the standing instruction, not built, and reversible (listed in docs/TODO.md).
G18: the accessibility table the re-audit scores against, with the rule that a requirement counts as shown only when its gated test passes and its interactive run is recorded.

ADR-0004

The step that builds the version 2 pipe server owns transport.get and printers.discover.
The tray tooltip sentence gains a pointer to G15.

Audit

Closed against merged code and tests: C24, C15, U20, U21, C9 (first half), W4, W25.
W26 partly closed: closed in the worker; the service's half waits for a runtime test in a later step.
New rows for nits earlier reviews left open: W90 to W95, C28 to C30, U23 to U27. W90 (nothing reconnects the realtime socket when a load balancer moves the HTTP side) is rated should-fix and scheduled in docs/TODO.md.
Three nits could not be recorded because no record of what they were survives (one each from pull requests 45, 50 and 53); the audit says so.

Session state and TODO

release/2.0 at 02f34c3 after pull requests 40 to 53, with the figures verified on that content.
"Needs the owner": whether 1.1.x and 1.2.x get the sign-in address fix; the real-printer check; the upgrade qualification; exception record EX-0016; the G19 decision.

Verification

Head 53f3110 on release/2.0 at 02f34c3. Documents only; no build or test was run. git diff --check clean, tables and links checked, source-header check passes.

Review (independent reviewer, three passes)

189b092: 1 blocker, 7 should-fix, 9 nits. The blocker was the release range above.
b72d652: Verdict on b72d652: 0 blockers, 1 should-fix, 2 nits.
53f3110, the head merged: Verdict on 53f3110: 0 blockers, 0 should-fix, 0 nits. (three wording corrections by the coordinator, confirmed by the reviewer).

Goes stale at later merges (to correct then)

When the server registry merges: W4's "statements only in JournalMigration.V1.cs", and line references in W4, W25, W26, C24, W91 to W93.
When the printer inventory change merges: the test and coverage figures, and the mentions of which changes are inside the real-printer check.
When the wizard's Server step merges: G20's "ManagementText is internal", and "not built" for G3, G5, G12.
When the identifier baseline merges: "none of those tests exists yet" under the accessibility definition.

🤖 Generated with Claude Code

Documentation only: the decisions taken while planning the connection wizard and server tree, audit closures and open review nits from pull requests 44 to 53, the session state, and what waits for the owner. It also corrects a merged statement about which releases log the sign-in address. **Correction of a merged statement** - Pull request 53's changelog entry, its description and audit row U22 said every release from 0.2.0-rc.1 to 1.2.1 writes the sign-in address to the companion's log when Windows cannot start the browser. By the code at the tags it is four releases: 1.1.0, 1.1.1, 1.2.0 and 1.2.1 (and `snapshot-2.0.0-rc.1`). The logging came with `aaa5051`, which only those tags contain. - The seven releases from 0.2.0-rc.1 to 1.0.1 start the browser the same way but catch the failure and show a fixed text. - Corrected here: the changelog entry, U22, `docs/TODO.md`, `docs/SESSION-STATE.md`. Corrected outside this change: the snapshot's release notes on Gitea, and a comment on pull request 53. **ADR-0005: "Resolved while planning the build, 2026-11-10/11"** - Decisions G1 to G21, each marked built, built in part or not built, with the code read to say so. - The sign-in provider rule is recorded as the merged contract enforces it: `servers.add` and `trust.set` state the provider the operator was shown; a missing value means the check named none; any disagreement is refused as `provider_changed`. - G7 (which control is the tree) and G11 (re-sending a credential after a refused bind) keep their conditions; neither is met yet. - G19: driving the layout host through UI Automation, and adding the layout tests to the solution when the interactive suites are built. The plan called the second half the owner's call; it is recorded as taken under the standing instruction, not built, and reversible (listed in `docs/TODO.md`). - G18: the accessibility table the re-audit scores against, with the rule that a requirement counts as shown only when its gated test passes and its interactive run is recorded. **ADR-0004** - The step that builds the version 2 pipe server owns `transport.get` and `printers.discover`. - The tray tooltip sentence gains a pointer to G15. **Audit** - Closed against merged code and tests: C24, C15, U20, U21, C9 (first half), W4, W25. - W26 partly closed: closed in the worker; the service's half waits for a runtime test in a later step. - New rows for nits earlier reviews left open: W90 to W95, C28 to C30, U23 to U27. W90 (nothing reconnects the realtime socket when a load balancer moves the HTTP side) is rated should-fix and scheduled in `docs/TODO.md`. - Three nits could not be recorded because no record of what they were survives (one each from pull requests 45, 50 and 53); the audit says so. **Session state and TODO** - `release/2.0` at `02f34c3` after pull requests 40 to 53, with the figures verified on that content. - "Needs the owner": whether 1.1.x and 1.2.x get the sign-in address fix; the real-printer check; the upgrade qualification; exception record EX-0016; the G19 decision. **Verification** - Head `53f3110` on `release/2.0` at `02f34c3`. Documents only; no build or test was run. `git diff --check` clean, tables and links checked, source-header check passes. **Review (independent reviewer, three passes)** - `189b092`: `1 blocker, 7 should-fix, 9 nits`. The blocker was the release range above. - `b72d652`: `Verdict on b72d652: 0 blockers, 1 should-fix, 2 nits.` - `53f3110`, the head merged: `Verdict on 53f3110: 0 blockers, 0 should-fix, 0 nits.` (three wording corrections by the coordinator, confirmed by the reviewer). **Goes stale at later merges (to correct then)** - When the server registry merges: W4's "statements only in `JournalMigration.V1.cs`", and line references in W4, W25, W26, C24, W91 to W93. - When the printer inventory change merges: the test and coverage figures, and the mentions of which changes are inside the real-printer check. - When the wizard's Server step merges: G20's "`ManagementText` is internal", and "not built" for G3, G5, G12. - When the identifier baseline merges: "none of those tests exists yet" under the accessibility definition. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

james.maes added 6 commits 2026-11-11 08:46:06 +00:00

docs(adr): record the decisions taken while planning the wizard and tree build 508436ff55

ADR-0005 gains a dated section, "Resolved while planning the build,
2026-11-10/11". It records each of the twenty-one gaps the plan for the
wizard stage listed (G1 to G21) as a decision: what was decided, what it
rules out, and whether merged code does it, read at 02f34c3. All were
taken under the standing instruction and are open to reversal by the
owner.

- G1 and G2 are recorded as the rule that was built in the contract by
  pull request 51, which is tighter than the wording of the plan: an
  absent provider origin states that the check named none, and it is
  compared in both directions.
- G7, G11 and G19 keep their conditions: the comparison of the two tree
  controls with Narrator and Accessibility Insights, what the service
  does with a session whose bind it refused, and the layout tests in
  the solution as a call for the owner.
- G18: the accessibility table of the plan is written in as the
  definition the re-audit scores against, with the result of a search
  of the standards at 707bbd33f0b9.
- The accepted limits of the plan are recorded.
- Six sentences of the record that a decision contradicted, or that
  said less than the rule as built, are corrected in place and listed.

ADR-0004 gains "Resolved while planning the wizard and tree build":
the step that builds the version 2 pipe server also builds
transport.get and printers.discover, the second with its single-flight
(G4). The sections "Resolved while building the server registry" and
"About printing" are not touched.

Documents only. Nothing was built or run.

Signed-off-by: James Maes <james@kof22.com>
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

docs(audit): bring nine statuses up to pull requests 44 to 53 and record what their reviews left open b996a90033

Statuses, each read against the code and the tests at 02f34c3:

- Closed, naming the pull request: C24, W25 and W26 (44), C15 (45),
  U20 and U21 (46), W4 (49). C9 keeps its second half open and names
  pull request 47 for the first.
- W26 is closed on the tests of the worker; that the runtime composes
  the same filter is by reading, and the row says so.
- W41 gains one case that is not redacted: a message over 8,192 units
  whose verifier name stands after the cut keeps its code.

New rows, in a table of their own, each read at 02f34c3 and rated as
the reviewer rated it, or "Not rated":

- W90 to W95: no reconnect when the load balancer moves the HTTP side
  to another node; three items from the journal schema step; the pipe
  client binding that only reading shows; one test that timed out
  under load.
- C28 to C30: what the reviews of pull requests 45, 47 and 48 left.
- U23 to U27: three items in the sign-out paths, what the review of
  pull request 46 left, and the older presentation types that log by
  joining strings.

Header: the progress paragraph has the figures of the head of pull
request 53; one sentence says that a number is an identifier and a gap
is not a missing row; "W74 and the rows after it" now reads "W74 to
W76".

Not written, and said so in the table's introduction: the last nit of
the review of pull request 50 and two items of the review of pull
request 45, because no record says what they are.

docs/ADMINISTRATION.md: a typo that pull request 46 added (a comma in
front of a semicolon) is corrected.

Documents only. Nothing was built or run.

Signed-off-by: James Maes <james@kof22.com>
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

docs: session state at 02f34c3 and what needs the owner 9620eab2c1

docs/SESSION-STATE.md:

- release/2.0 is at 02f34c3 after twenty-nine pull requests; the
  fourteen merged since the snapshot (40 to 53) are listed one line
  each.
- The figures verified on that content are those the description of
  pull request 53 records for its reviewed head 0b9102e, whose tree is
  the tree of 02f34c3: 1,174 Core, 2,321 Windows runtime and command
  line and 537 presentation tests, none failed; Core 100% of 2,311
  lines, presentation 100% of 1,373, command line 100% of 120, Windows
  96.08% (4,540 of 4,725).
- No record shows anything from release/2.0 run elevated, installed or
  printed. The snapshot was not installed, not UI-tested, and nothing
  was printed with it.
- What is still to build, from the two plans, and that the upgrade
  qualification, the two checks on real printers and the re-audit have
  not been run.

docs/TODO.md gains "Needs the owner": whether a 1.2.x release gets the
fix for the sign-in address in the log (audit U22), with the rule that
a 1.2.2 is not released without it; the check on real printers, which
is due; the administrator prompt of the upgrade qualification; and the
approval of the corrected exception record EX-0016.

Documents only. Nothing was built or run.

Signed-off-by: James Maes <james@kof22.com>
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

docs(adr): say which wizard decisions the plan marked for the owner 189b092c33

The section "Resolved while planning the build" now says that the plan
marked G1, G2, G4, G5, G7, G11 and G19 as needing the owner or the
service steps, that they were taken under the standing instruction like
the rest, and which of them keep a condition. The sentence about which
audit findings count against the accessibility criterion gives the
plan as its source and leaves the judgement to the reviewer of the
re-audit.

Documents only. Nothing was built or run.

Signed-off-by: James Maes <james@kof22.com>
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

docs: four releases log the sign-in address, not eleven; answer the review of 189b092 b72d652922

The independent review of 189b092 found one statement that was wrong
and merged, and sixteen smaller things. Documents only; nothing was
built or run.

Which releases write the sign-in address to the log (corrects a merged
changelog entry and a merged audit row):

- By the code at the tags, a failed browser start logs and shows the
  provider's sign-in address in v1.1.0, v1.1.1, v1.2.0 and v1.2.1 and
  in snapshot-2.0.0-rc.1. The log call and the "Reason:" text came
  with aaa5051 (git tag --contains aaa5051 gives exactly those five).
  The seven tags from v0.2.0-rc.1 to v1.0.1 start the browser the same
  way, since 2c1c5f1, but catch the failure and show a fixed text; the
  catch is identical at all seven and the file has no log call there.
- CHANGELOG.md, the entry "A failed browser start no longer writes the
  sign-in address ...", merged with pull request 53, said "Every
  release from 0.2.0-rc.1 to 1.2.1 has the sign-in case". It now names
  the four, and says what the seven do. This is the only change to the
  changelog.
- Audit U22 said "the eleven version tags" and "introduced by
  2c1c5f1". Corrected, with the date and what the row said before.
- docs/TODO.md, the owner's first item, had the same range and an
  unsupported clause about the nonce, which every tag sends.
- docs/SESSION-STATE.md says where the wrong range stood and that the
  snapshot's release notes and a comment on pull request 53 were
  corrected by the integrator's record, which was not read on Gitea.

Audit:

- W26 is partly closed, not closed: closed in the worker by pull
  request 44, open in the service's pipeline, where the plan's runtime
  test does not exist yet. Only unsharing is said to be settled by the
  filter.
- C8: the second line reference was AgentRuntime.RealtimeStatus.cs:136
  and is :184 since db45420; the row also says that a frame from the
  socket cannot have a null payload.
- W90 is rated Should-fix by the review of this change, says what the
  runtime does after a refused claim (traced by reading), and has an
  item under "Open for 2.0" in docs/TODO.md.
- C28 item 4 names the reader's summary in PrintingRealtimeChannel.cs
  that is untrue for a frame with an unreadable member name, as found
  by the review of this change. C30 item 1 gains the consequence its
  review stated. U25 says what QqqProviderLogout.Matches asks for.
- The sentence about row numbers no longer contradicts itself.

ADR-0005 and ADR-0004:

- G11: a new pairing needs a new sign-in again, which is the rule in
  force; the sentence that it falls away is tied to G11's condition,
  which is not met.
- G19: adding the layout tests to the solution was the plan's "owner's
  call". It is recorded as taken under the standing instruction, not
  built, for the step that builds the interactive suites, and open to
  the owner's reversal before then. docs/TODO.md lists it for the
  owner as a fifth item, and the session state says so.
- A requirement of the accessibility definition counts as shown only
  when its gated test passes and its interactive run is recorded for a
  build of the same commit; what each run must observe is left to the
  step that makes the runs.
- What the plan listed as needing a decision, which refusal the root
  certificate and the exception are for, the fourth smaller point, and
  how the in-place corrections are marked.
- ADR-0004, "Companion": the tooltip sentence carries a pointer to
  G15.

Signed-off-by: James Maes <james@kof22.com>
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>

docs(audit): say which numbers were set aside for what, and where the second read of a command stands 53f3110a0b

Three wording corrections asked for by the confirmation pass of the review of
this documentation change:

- The numbering sentence said W96 upward and W100 upward were set aside for two
  more changes still in review. W96 upward is the fix round of one of the two
  changes that hold W82 to W89, and W100 upward is a third change that was being
  written.
- The paragraph on statuses says that two sentences were added to C8 beside the
  corrected line reference.
- C8 placed the second deserialization of a queued command under 'before that
  validation runs'. Only the first is before it.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Signed-off-by: James Maes <james@kof22.com>

james.maes merged commit 4c9b3686d3 into release/2.0

2026-11-11 08:46:18 +00:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: QRun-IO/qqq-print-agent#54