Skip to content

docs: the operation coverage table names its profile - #56

Merged
jonathanstokes merged 1 commit into
mainfrom
docs/operation-coverage-per-profile
Sep 9, 2026
Merged

jonathanstokes merged 1 commit into
mainfrom
docs/operation-coverage-per-profile

Conversation

@jonathanstokes

@jonathanstokes jonathanstokes commented Sep 9, 2026 •

Copy link
Copy Markdown
Contributor

Summary

The ## Operation coverage header in docs/protocol.md still promised that every operation was exercised on firmware d14e, CorOS 4.0.1. That stopped being the rule when ADR-0020 landed.

The note at the top of the same file already says a statement names its profile, and that an observation from another profile is written beside the 4.0.1 record and never in its place. The table's own header contradicted that. So a row measured on another profile had nowhere to live except in place of the baseline.

This came up on PR #47, where a contributor hit exactly that and rewrote the header to admit a CorOS 4.1.0 row. I asked them to revert it, because the header is mine to fix rather than theirs.

The change also documents captured only, which the table already uses once and which a row with no independent read-back needs.

docs/STEERING.md needs no change here: it already states the per-profile rule.

Follow-up

Once #47 merges, the 4.0.1 chunk behaviour I measured for create_local_backup goes in beside its 4.1.0 row. On 4.0.1 a 130,178-character backup arrives as ONE message, so the 150,000 figure looks like a chunk ceiling rather than a fixed chunk size.

Hardware

Docs-only, and the suite ran anyway on the final commit c63f891, on Quad Cortex, CorOS 4.0.1 (app_fw d14e):

============== operations on QuadCortex (CorOS 4.0.1, MAINTAINER) ==============
tests: 95 passed, 0 failed, 0 errored, 4 skipped; operations measured: 16 of 16 claimed
passed, not VERIFIED:                   [] (0)   <- candidates to add
VERIFIED and claimed by a test, failed: [] (0)   <- regressions by name
================== 95 passed, 4 skipped in 188.31s (0:03:08) ===================

Offline suite 2698 passed, 3 skipped. mypy clean on 46 source files.

One thing worth recording. An earlier run of this same commit failed two tests, test_the_connect_burst_warms_the_cache and test_the_burst_delivered_every_entry_the_preset_surface_reads, with "the burst delivered nothing for preset".

Those two check something the unit does about ten seconds after you connect: it sends its current state all at once, which preset is loaded, which scene is active, and so on. The model layer fills its cache from that. The tests check the cache really got filled, and this time it had not.

It is not this change. Both tests pass when re-run, both pass on unmodified main, and this branch is 13 lines of Markdown, which cannot affect what the unit sends. So the unit is sometimes slow to send it. Written up in issue #57.

…f the file

The header still promised that every operation was exercised on firmware d14e,
CorOS 4.0.1, which stopped being the rule when ADR-0020 landed. The note at the
top of this file already says a statement names its profile and that an
observation from another profile sits beside the 4.0.1 record. The table's own
header contradicted that, so a row measured on another profile had nowhere to
live except in place of the baseline.

Also documents `captured only`, which the table already uses once and which a
row with no independent read-back needs.

STEERING.md needs no change: it already states the per-profile rule.
@jonathanstokes
jonathanstokes marked this pull request as ready for review September 9, 2026 19:45
@jonathanstokes
jonathanstokes merged commit 0476694 into main Sep 9, 2026
5 checks passed
@jonathanstokes
jonathanstokes deleted the docs/operation-coverage-per-profile branch September 9, 2026 19:46
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