Skip to content

docs: cover from the edge, and keep the coverage floor - #911

Merged
reardencode merged 2 commits into
masterfrom
docs/cover-from-the-edge
Oct 5, 2026
Merged

reardencode merged 2 commits into
masterfrom
docs/cover-from-the-edge

Conversation

@rearden-grok

@rearden-grok rearden-grok Bot commented Oct 4, 2026 •

Copy link
Copy Markdown
Contributor

Summary

The 92% production line floor and the CRAP gate stay. This updates the docs agents read so those gates are held by covering a line from a surface journey, or by deleting a branch no peer, client, or operator can trigger. A private-helper test is not how a red coverage or CRAP job goes green.

TESTING.md owns the rule. CONTRIBUTING.md, AGENTS.md, and docs/how-we-plan.md point at it and do not restate the procedure. Catalog cells that said "keep this unit" now name the result the journey asserts. No CI scripts change.

The same branch also fixes client_that_stops_reading_is_dropped_at_the_write_deadline, which failed the required test job twice. A 200ms client send is not a full window: on a busy runner the test stopped writing while the session was idle, so the write deadline never started. Linux autotunes both SO_SNDBUF and the peer receive window, and write() keeps completing until several MiB are queued or ACKed. The test now pins both buffers and keeps sending until that stalled write resets the socket.

Test plan

  • cargo test -p rbitcoin-sv2 --lib and clippy -D warnings for that crate. The deadline test passed 8 times locally.
  • Required checks: qc, test, windows, macos, coverage.

The 92% line floor and the CRAP gate stay. A lower-level test comes out
only once a surface journey already hits those lines. A branch no peer,
client, or operator can trigger is deleted instead of given a private
helper test. Say that in TESTING.md, and point CONTRIBUTING, AGENTS, and
the planning cycle at that rule.
A 200ms client send is not a full TCP window. On a busy runner that
timeout is scheduler delay; the test then stops writing, the session
sits idle, and the write deadline never starts. Linux also autotunes
SO_SNDBUF and the peer receive window, so write() keeps completing
until several MiB have been queued or ACKed. Pin both buffers and keep
sending until the stalled write resets the socket.
@reardencode
reardencode merged commit f93930e into master Oct 5, 2026
18 checks passed
@rearden-grok rearden-grok Bot mentioned this pull request Oct 5, 2026
3 tasks done
@rearden-grok
rearden-grok Bot deleted the docs/cover-from-the-edge branch October 5, 2026 03:55
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