Skip to content

Add ordering-switch completeness guardrail to M1 - #506

Merged
sysread merged 1 commit into
mainfrom
claude/conversation-forking-8ntvpq
Aug 19, 2026
Merged

Add ordering-switch completeness guardrail to M1#506
sysread merged 1 commit into
mainfrom
claude/conversation-forking-8ntvpq

Conversation

@sysread

@sysread sysread commented Aug 19, 2026

Copy link
Copy Markdown
Owner

Doc-only amendment from the final review round's sign-off: M1 gains a completeness proof for the created_at-to-position ordering switch - a grep-based permanent guardrail test with an allowlist of the legitimate cross-thread orderings (same pattern as the style/markdownlint guardrail tests), upgraded from the review's one-time manual grep because the bug it guards against is invisible on fresh data (the two orderings coincide until a recovery row lands with a fractional position), so a later-added created_at reader would regress silently without it.


Generated by Claude Code

The final review round signed off on the plan with one addition: a
grep proving no message-reading path still orders by created_at
after the position switch. Folded in as a permanent guardrail test
with an allowlist of the legitimate cross-thread orderings, rather
than a one-time manual check - the bug it guards against is
invisible on fresh data (the two orderings coincide until a
recovery row lands with a fractional position), so a straggler or
a later-added query would pass every demo and misorder exactly the
transcripts that needed healing.
@sysread
sysread merged commit ad11d29 into main Aug 19, 2026
1 check passed
@sysread
sysread deleted the claude/conversation-forking-8ntvpq branch August 19, 2026 21:37
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.

2 participants