Skip to content

docs(plan): record the M1.6 application check as run, and close M1 - #41

Merged
aenawi merged 1 commit into
mainfrom
m1-application-check
Sep 5, 2026
Merged

aenawi merged 1 commit into
mainfrom
m1-application-check

Conversation

@aenawi

@aenawi aenawi commented Sep 5, 2026

Copy link
Copy Markdown
Owner

Docs only. No code changed.

M1's last open box was the one no test could tick — docs/PLAN.md §1.6:

Application check, NOT RUN. The issue's last criterion is a person opening the repaired deck and seeing the labels laid out right-to-left. That cannot be proven by a test, and it has not been run. Schema validity is not the same claim.

It has now been run

The maintainer opened a repaired torture.pptx in PowerPoint:

  • no repair prompt
  • no error
  • text laid out right-to-left

The item is ticked with the date it was run. The *Acceptance:* line above it drops its "met for the machine half" hedge, since both halves are now met.

The check stays human

It is recorded here rather than automated, because the thing being verified is what a rendering engine does with a part mirsam wrote — which is exactly what a schema validator cannot answer. The chart c:txPr was the first part the writer creates from nothing rather than edits in place, so the note added with the item says the next writer to do that should be run the same way before its own box is ticked. The precedent travels with the plan instead of living in a merged PR.

Where the plan stands

M0 through M5 are clean. M6 (PDF, read-only) and M7 (Distribution) are what remain.

The last open box in M1 was the one no test could tick: a person opening
the repaired deck and seeing the chart's category labels laid out
right-to-left. Schema validity is not that claim, and the item said so.

The maintainer ran it. A repaired `torture.pptx` opened in PowerPoint
with no repair prompt, no error, and the text laid out right-to-left. The
item is ticked with the date, and the acceptance line above it drops its
"met for the machine half" hedge, because both halves are now met.

The check stays human and is recorded rather than automated. The note
added with it says the next writer that creates a part from nothing —
the chart `c:txPr` was the first — should be run the same way before its
own item is ticked, so the precedent travels with the plan instead of
living in one merged pull request.

M0 through M5 are now clean; M6 and M7 are what remain.
@aenawi
aenawi merged commit 5fdd076 into main Sep 5, 2026
7 checks passed
@aenawi
aenawi deleted the m1-application-check branch September 5, 2026 15:35
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.

1 participant