Skip to content

Fix three stale references found auditing today's own changes - #17

Merged
oroderico merged 1 commit into
mainfrom
fix-stale-comments-and-docs
Aug 16, 2026
Merged

Fix three stale references found auditing today's own changes#17
oroderico merged 1 commit into
mainfrom
fix-stale-comments-and-docs

Conversation

@oroderico

Copy link
Copy Markdown
Owner

Documentation and a self-test comment had drifted from the code they describe. None of these are from today's PRs — the oldest predates them by a week — but all three surfaced while checking today's work for the same failure mode, and none had been caught since.

README's Controls section quoted footer text that no longer exists

L/R move BOTH select, BOTH continue L/R back and BOTH extend L/R erase — the board's two buttons sit one above the other, not side by side, and the code has said Up/Down since 5e23ce2 (8 Aug). That commit updated NAV_FOOTER, the timeout warning and the About screen's own prose, but never the top-level README describing the same buttons.

Worse than a wrong word: BOTH continue L/R back doesn't name any real footer any more. That screen moved to the chrome's tick, which draws no bar at all.

A self-test comment claimed a copy that no longer exists

dice_screen_hints_clear_the_bar said it tested "every hint the app actually shows, copied from collect_entropy()" — but all three hints, including the "Looks good - generate?" it quoted, migrated to nav_dice_confirm (83b7e46, 15 Aug) and are no longer passed to the plain renderer this test drives.

The check itself stayed valid despite that: seedtool_render_screen's two-line layout and seedtool_render_nav_text's share the same y=39/65 — which is exactly what nav_chrome_bands_do_not_collide's own comment leans on when it skips re-proving the 90..103 band. Fixed the hint text to match what nav_dice_confirm shows today, and the comment to explain why checking it through the plain renderer is still meaningful, rather than claiming a copy that no longer exists.

The audit skill pointed at a revoked ceiling and one contributor's machine

.claude/skills/audit/SKILL.md told an auditor to check the binary against a 295 KiB ceiling; tools/audit_origo_elf.py's default is 512 KiB, raised weeks ago and explained in that file's own comment. Read literally, today's 303920-byte build would score as over the limit instead of at 58% of it.

The skill also named one contributor's home directory and asserted their default IDF had drifted to 6.0.2 — a fact about a specific machine, stated as if it held everywhere. Replaced with placeholders and a pointer to dependencies.lock's pinned version, so the skill stops being accurate for exactly one filesystem.

Left alone

README's "Going back" section still describes a Back row at the end of every menu, which the chrome's arrow replaced weeks before this. Bigger than a wording fix — it needs rewriting for the current model, and possibly "Lists and scrolling" beside it — so it is flagged here rather than folded into this diff.

Verification

Host self-test and the 52 Python tests. No firmware change; nothing here touches main/ or the ELF.

Documentation and a self-test comment had drifted from the code they
describe. None of these are from today's PRs - the oldest predates them
by a week - but all three surfaced while checking today's work for the
same failure mode, and none had been caught since.

README's Controls section still quoted `L/R move   BOTH select`,
`BOTH continue   L/R back` and `BOTH extend   L/R erase` as on-screen
footer text. The board's two buttons sit one above the other, not side
by side, and the code has said "Up/Down" since 5e23ce2 (8 Aug) - that
commit updated NAV_FOOTER, the timeout warning and the About screen's
own prose, but never the top-level README describing the same buttons.
Worse than a wrong word: `BOTH continue L/R back` doesn't name any real
footer any more, since that screen moved to the chrome's tick, which
draws no bar at all.

host/origo_simulator.c's dice_screen_hints_clear_the_bar claimed to test
"every hint the app actually shows, copied from collect_entropy()" - but
all three hints, including the "Looks good - generate?" it quoted,
migrated to nav_dice_confirm (83b7e46, 15 Aug) and are no longer passed
to the plain renderer this test drives. The check itself stayed valid
despite that: seedtool_render_screen's two-line layout and
seedtool_render_nav_text's share the same y=39/65, which is exactly what
nav_chrome_bands_do_not_collide's own comment leans on when it skips
re-proving the 90..103 band. Fixed the hint text to match what
nav_dice_confirm shows today, and the comment to say why checking it
through the plain renderer is still meaningful rather than claiming a
copy that no longer exists.

.claude/skills/audit/SKILL.md told an auditor to check the binary
against a 295 KiB ceiling; tools/audit_origo_elf.py's default is 512
KiB, raised weeks ago and explained in that file's own comment. Read
literally, today's 303920-byte build would score as over the limit
instead of at 58% of it. The skill also named one contributor's home
directory and asserted their default IDF had drifted to 6.0.2 - a fact
about a specific machine stated as if it held everywhere. Replaced with
placeholders and a pointer to dependencies.lock's pinned version, so the
skill stops being accurate for exactly one filesystem.

Left alone: README's "Going back" section still describes a "Back" row
at the end of every menu, which the chrome's arrow replaced weeks
before this. Bigger than a wording fix - it needs rewriting for the
current model, and possibly "Lists and scrolling" beside it - so it is
flagged rather than folded into this diff.

Verified with the host self-test and the 52 Python tests. No firmware
change; nothing here touches main/ or the ELF.
@oroderico
oroderico merged commit 373f393 into main Aug 16, 2026
3 checks passed
@oroderico
oroderico deleted the fix-stale-comments-and-docs branch August 16, 2026 17:08
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