Repository navigation
mtcollins1 socket 1: record operator-reported pin damage (2026-10-03) - #13079
Conversation
… 2026-10-03 CPU removal Paste fell on the socket-1 pins during the CPU removal for the one-socket test, and a microfiber cloth used to clean it probably bent some. A typed OperatorByEyeReport row with the account and the consequence (any later socket-1 test must inspect the socket first; a socket-1 failure after this date is not evidence about the pre-damage fault), rendered in the unit's history block and asserted by its witness. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
…a typed evidence gate read by the renderers
Review 74527: the history label re-typed the date the report carries, and the 'inspect first'
consequence was prose nothing read. socket_evidence_gate derives SocketEvidenceRequiresInspectionSince
{ socket, since } from the report's date; the history line and every OPEN socket-1 hypothesis render
it, and date_text formats the record's date. Witness: a report dated 2027-01-09 renders that date
and gate, not 2026-10-03.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
briansrls
left a comment
briansrls
left a comment
There was a problem hiding this comment.
Approved exact head 2b8f175564.
The review-74527 defects are resolved. The rendered date is derived from the report's observed_on; the consequence is no longer a free-form field but MtCollins1SocketEvidenceGate, derived by socket_evidence_gate; both the history line and every HypothesisOpen rendering consume that gate. The 2027-01-09 control proves the date and gate are not hard-coded to the production datum.
The evidence standing also remains honest: this is an operator report of probable pin damage with no magnified inspection, and the gate applies prospectively without rewriting the pre-damage boot receipts or their existing hypothesis standings.
Two non-blocking presentation cleanups:
SocketEvidenceRequiresInspectionSincerenders as evidence "after" the date; "on or after" would match the inclusiveSincesemantics and the date-only precision more exactly.- the PR body still says 15 orientation claims; the current stated result is 16/16.
Merge after the exact-head required workflow is green.
1 participant
Records the operator's report that socket 1's pins were probably bent during the 2026-10-03 CPU removal: paste fell on the pins, and a microfiber cloth used to clean it snagged them. It is
MtCollins1SocketDamageReportingunbc.machine_intake_mtcollins1_socket1_investigation_observation, rendered in the unit history bymtcollins1_physical_orientationhistory_blockand asserted byw_slot_map_carries_cpu_facts_and_hypothesis_standing. Consequence recorded: any later socket-1 test must inspect the socket first, and a socket-1 failure after this date is not evidence about the pre-damage fault. Socket 1 has been empty since. All 15 orientation witness claims PASS (BuildBuddy).🤖 Generated with Claude Code