Skip to content

fix: doctor keeps a CRLF .gitignore intact, the HTML journey report shows resolved seed values, the progress report names State words it does not know (#668, #653) - #761

Merged
MendixMau merged 9 commits into
masterfrom
claude/loving-gates-hauc1n-bugs4a
Nov 15, 2026
Merged

MendixMau merged 9 commits into
masterfrom
claude/loving-gates-hauc1n-bugs4a

Conversation

@MendixMau

Copy link
Copy Markdown
Owner

Direct lane

What changed and why (one paragraph): Three reporting fixes. (A) bin/doctor.sh appends to a project's .gitignore through the shared helper, so a file with Windows line endings keeps them and gets no duplicate blocks — the last script with that flaw. (B) The HTML journey report shows resolved seed values ("Open SO-1001") where it showed {{order}}; only the renderer changed, report.json and every join key are byte-identical (closes #668). (C) progress-report.py no longer counts a State word it does not define as "not built" silently: it gets its own "unrecognised state" count and each such row is named; totals still add up; report only, no gate, no vocabulary change, as the issue's triage chose (closes #653).

Row Answer
Process none
Dependencies doctor.sh (sources bin/lib/gitignore-ensure.sh), report-render.js, progress-report.py; journey-runner.js exports walks for fixtures
Context cost none: no routed or always-loaded file touched (baseline 66,196 on base and head)
Behaviour n/a: scripts only
Runtime n/a: no check added

Measured, base vs head: (C) over every build-plan in the repo's fixtures (15 plans / 71 rows through build-plan-status.sh --json, 8 plans / 49 rows through the markdown reader) 0 rows change classification; (B) the same captured walk rendered by both versions differs in 28 HTML lines, all displayed text, report.json identical; (A) an LF .gitignore is byte-identical after the old and the new doctor.

For an existing project after sync: a CRLF .gitignore keeps its endings; doctor's "(added … to .gitignore)" note prints only when a line was actually written; the report pages show resolved values and name unrecognised State words. No verdict or exit code changes.

Field evidence — (B) golden input is a capture of the real runner in headless Chromium against a local stand-in page. (C) the three State words are verbatim from the issue; no repo fixture carries one. (A) CRLF input is the reporter's shape from the earlier fix; one doctor site (property-sets receipt) runs only with a real mxcli and was not reproduced.

  • No client data anywhere in the diff
  • Size: 12 files, 396 lines including fixtures
  • Test tier: scoped fixtures — test-gitignore-crlf.sh 31/31, test-report-seeds-668.sh 9/9 (3/9 against master's renderer), test-progress-report.sh 127/127 and the other progress-report fixtures
  • CHANGELOG.md lines appended, credited

Generated by Claude Code

…ames and check labels (#668, report half)

report-render.js resolves {{seed}} placeholders for display only, on a copy, from walks[].seeds; stored names, report.json and every join key stay raw. Unresolved placeholders stay visible.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013jQvaonym9V1wKKfAEPmRr
…a mxtk_gitignore_ensure

doctor.sh wrote LF lines into a CRLF .gitignore (mixed endings) at four sites.
LF files get the same bytes as before; the 'added ...' note prints only when a
line was written. Fixture: doctor cases (CRLF, LF byte-identity, rerun).

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013jQvaonym9V1wKKfAEPmRr
…not counted as todo (#653)

classify_state returns 'unread' for a State word the table does not define;
the row stays in the totals, gets its own count and an 'unrecognised state:
<word>' note. Blank State stays todo. Display only.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013jQvaonym9V1wKKfAEPmRr
…68-report' into claude/loving-gates-hauc1n-bugs4a

# Conflicts:
#	CHANGELOG.md
…53' into claude/loving-gates-hauc1n-bugs4a

# Conflicts:
#	CHANGELOG.md
…; T18 gaps; unrecognised-state styling (#653)

A State cell holding only -, em/en dash, [ ] or TBD is an empty-cell marker like a blank cell, not a state word. Claims outside the vocabulary (x, check mark, yes, ...) stay unread and named. Legend and r-unread row use the page's --bad token. test-gitignore-crlf: second doctor run prints no '(added' note.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013jQvaonym9V1wKKfAEPmRr
…a negated-entry case

On a rerun git check-ignore says ignored, so the (added block is never reached and the assertion could not fail. Now every entry is present but a later negation makes git report it not ignored: doctor reaches each block, the helper writes nothing, no note, bytes unchanged. Fails when the note is made unconditional.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013jQvaonym9V1wKKfAEPmRr
@MendixMau
MendixMau merged commit 46624b1 into master Nov 15, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

2 participants