Skip to content

process(inbox): triage sweep 1 — 12 drops deleted, 20 promoted, 11 kept - #811

Open
MendixMau wants to merge 2 commits into
masterfrom
triage/inbox-sweep-1
Open

MendixMau wants to merge 2 commits into
masterfrom
triage/inbox-sweep-1

Conversation

@MendixMau

@MendixMau MendixMau commented Nov 20, 2026 •

Copy link
Copy Markdown
Owner

Direct lane

What changed and why (one paragraph): contrib/inbox/ had 43 drops (Aug–Nov 2026), and CONTRIBUTING.md calls the inbox "a queue, never an archive". This PR is triage sweep 1. Every drop got one row: its claims, whether each claim is already on master (cited file:line), stale, or new, and a verdict.

  • 12 drops deleted. Master already carries their content, or later fixes superseded it. Examples: exec.sh's crash-after-write rollback, the Stage-7 table scoping, the umlaut ledger fix, and the 0.26 round-trip report U1–U53.
  • 20 drops promoted, and deleted in the same commit:
    • seven ## BUG-DRAFT-* entries plus a BUG-39 addendum in bug-logs/mxcli-bugs.md;
    • an addendum A1–A19 to the not-yet-filed 0.26 round-trip report;
    • 3–5 line notes in on-demand skills;
    • the project-bin/mxpatch.py docstring.
  • 11 drops kept. Each needs a field run, a captured golden input or a maintainer decision.

Related cleanup:

  • The seven 2026-09-27 pending upstream drafts cited a deleted inbox file; their Source: line now describes the source instead.
  • Promoted text uses generic names.

Not in this PR:

  • The two drops in PR contrib/inbox: learnings from a requirements-driven full build #804 both triage to DELETE, and their filenames carry a project repo name.
  • A bug found while triaging: project-tests/e2e/helpers.js:633-635 fullPageShot uses fullPage: true, which crops an Atlas page at the viewport. It deserves its own fix PR with a field capture.

Impact

Row Answer
Process none. The inbox shrinks from 43 to 11 drops.
Dependencies Changes to existing files: bug-logs/mxcli-bugs.md (append-only, plus the 3-line BUG-39 addendum), the round-trip report (append-only), and project-bin/mxpatch.py (docstring only, ast.parse clean). No _common.sh, exec.sh, skill-routing.tsv or mdl-rule-kinds.tsv change.
Context cost none for always-loaded text: baseline is 65568 → 65568 (render-routing.sh --check). mdl-rules.sh --check is in sync. On-demand skills and the docstring grow by 936 words, and bug-logs/ by 1,586.
Behaviour NOT RUN: these are short facts in on-demand files, each with its measurement. No existing rule text was reworded.
Runtime n/a

Time cost: none · 0 s · can force rework: n · offset: n/a

Field evidence: none new. Every promoted line carries the field run, date and versions its drop recorded (mxcli v0.21–v0.27, Mendix 10.24–11.15). Defects from before v0.25 are marked "not retested".

  • No client data. The leak guard is clean, but it ran with LEAKGUARD_ALLOW_NO_DENYLIST=1, so the name check was not exercised. CI's denylist will confirm.
  • Size cap: exceeded, 54 files, +148 / −2293. Nearly all of it is inbox deletions; the additions are 148 lines over 22 files. Splitting would separate a promotion from its deletion, which CONTRIBUTING.md asks to keep in one commit. Maintainer's call.
  • Test tier: T0. check-pr-discipline, render-routing --check, mdl-rules --check, check-scripts (145/145 + 30/30), check-portability and check-no-private-citations are all clean. No fixture run: the only code file changed in its docstring.
  • CHANGELOG.md line appended, crediting "inbox sweep, contributors named in each promoted entry"
  • New bug entries headed ## BUG-DRAFT-<slug>:

Review: pending independent review


Generated by Claude Code

13 drops were already on master or superseded and are deleted. 19 are
promoted and deleted in this commit: seven BUG-DRAFT entries and a BUG-39
addendum in bug-logs/mxcli-bugs.md, an addendum A1-A19 to the not-filed
0.26 round-trip report, and short notes in on-demand skills
(fixture-seeding, learned-constants-and-secrets, rest-integration,
e2e-harness-base, learned-workflow-patterns, learned-page-patterns,
learned-file-upload-widget, existing-app-change, app-analysis,
doctor-triage, mdl-cookbook-microflows) plus the mxpatch.py docstring.
The seven 2026-09-27 pending upstream drafts no longer cite a deleted
inbox file. No baseline skill text changed (render-routing --check:
65568 / 80000 before and after).

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01HXpQSjXefSNicGGqP4aiSC
@MendixMau

Copy link
Copy Markdown
Owner Author

Review: MERGE-AFTER-FIX. One DELETE verdict loses content: the 9.24 runtime-boot facts in 2026-10-12-increment-4-seam-probe-findings.md are not on master or on the branch.

Value: Clears 32 of the 43 drops. Each promotion is deleted in the same commit, and promoted text goes to on-demand files.

DELETE spot-check, 6 rows: 5 are OK, with the content on master at the cited lines. increment-4-seam-probe is partly missing. These facts appear nowhere outside the inbox:

  • MX_INSTALL_PATH;
  • every constant passed in update_configuration;
  • mxbuild 9.24 --java-home=<jdk17>;
  • the 10.24→11.13 REST call needing no mappings;
  • the Java API edits.

PROMOTE spot-check: all 7 rows are faithful, with matching numbers. One gap: the retry-loop row leaves out the drop's limit that the repeat path is not exercised. No client names; the added names are generic.

Regression risk: Nothing executable changed (mxpatch.py docstring only). The 11 changed skills are all on demand, and mdl-rules.sh --check is in sync. No test, bin or .github file references the deleted files. No conflict.

Correctness: The seven ## BUG-DRAFT-<slug>: entries follow the convention, carry a NOT-YET-FILED banner and are not already covered. No slug is duplicated and no BUG number is used. No promoted text cites an inbox file. The PR exceeds the size cap, as it states; that's the maintainer's call.

Context cost: 65,568 → 65,568 (measured).

Time cost: none.

Exact fixes required:

  1. Move increment-4-seam-probe from DELETE to PROMOTE: add 3–4 lines on the 9.x runtime boot to skills/decompose-app.md near :32, and fix the triage table row. Or KEEP the file.
  2. Optional: add the "repeat path not exercised" limit to the retry-loop cookbook row.

Generated by Claude Code

…the retry loop's limit

Independent review (MERGE-AFTER-FIX): the increment-4 seam-probe drop was marked DELETE, but
its 9.24 runtime-boot facts were nowhere outside the inbox. Moved to PROMOTE: five lines
in decompose-app.md Stage 0 item 3 (Java 11 plus MX_INSTALL_PATH, every constant in
update_configuration, mxbuild 9.24 with --java-home=<jdk17>, the Java API edits on the 10.24
copy), worded and versioned as in the drop. The retry-loop cookbook row now carries the
drop's limit: the loop's repeat path was not exercised. Counts are now 12 deleted and 20
promoted (32 still leave the queue). decompose-app and mdl-cookbook-microflows are on demand:
baseline 65568 / 80000 unchanged; mdl-rules --check in sync.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01HXpQSjXefSNicGGqP4aiSC
@MendixMau MendixMau changed the title process(inbox): triage sweep 1 — 13 drops deleted, 19 promoted, 11 kept process(inbox): triage sweep 1 — 12 drops deleted, 20 promoted, 11 kept Nov 20, 2026
@MendixMau

Copy link
Copy Markdown
Owner Author

Review fixes pushed in c08bcd6.

  1. The increment-4-seam-probe drop moved from DELETE to PROMOTE. skills/decompose-app.md Stage 0 item 3 gains 5 lines, in the drop's own wording:
    • the 9.24.2 runtime needs Java 11 (17 is refused) and MX_INSTALL_PATH;
    • pass every constant in update_configuration, or the run fails on MendixSSO.TokenEncryptionKey;
    • mxbuild 9.24 builds with --java-home=<jdk17>;
    • the Java API edits on the 10.24 copy (final generated parameters, rollbackTransAction → rollbackTransaction).
  2. The retry-loop cookbook row now carries the drop's limit: the repeat path was not exercised.

Corrected counts: 12 deleted, 20 promoted, 11 kept. This supersedes the body's "13 / 19"; decompose-app.md joins the promoted skills.

render-routing is in sync, and the baseline stays at 65,568 (both changed skills are on demand). mdl-rules --check is in sync. Needs a re-review.


Generated by Claude Code

@MendixMau

Copy link
Copy Markdown
Owner Author

Review: MERGE — re-review of c08bcd6: the 9.24 runtime-boot facts are promoted into skills/decompose-app.md:35-39, and the retry-loop row keeps the drop's limit.
Value: I compared the text line by line with origin/master:contrib/inbox/2026-10-12-increment-4-seam-probe-findings.md:32-47. Every fact matches, versions included:

  • Java 11 (17 refused, Java 11 is required);
  • MX_INSTALL_PATH / Error finding mx install path.;
  • every constant in update_configuration / MendixSSO.TokenEncryptionKey;
  • mxbuild 9.24 --java-home=<jdk17>;
  • final DeepClone parameters and rollbackTransAction → rollbackTransaction.

The drop's REST point ("10.24→11.13 needed no mappings") is already covered at skills/mendix-service-seam.md:47. Only "start takes 10 s on HSQLDB" is left out, which is fine. The mdl-cookbook-microflows.md:578 row now ends with "Limit: the loop's repeat path was not exercised (the whole-loop test's stub always succeeds)", which matches the drop.
Regression risk: None. Both files are on demand. mdl-rules.sh --check reports skills/generated/ in sync, and render-routing.sh --check is in sync. The inbox holds 12 files (11 KEEP plus TEMPLATE).
Correctness: The CHANGELOG counts were corrected (12 deleted, 20 promoted). Before squash: the PR body still says "13 drops deleted" and "19 drops promoted". Update it.
Context cost: 65,568 → 65,568 (measured).
Time cost: none.
CI / merge: checks was in progress, authoring-audit passed. Merges cleanly on origin/master; the CHANGELOG line conflicts with sibling PRs, which is trivial.


Generated by Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants