Used the provided stock client as an oracle to verify the event read path. The
capture-event harness returns 50 real events, and the instrument-grpc-nonstream rewrite
captured the exact GetNextEventQueryResultBuffer.result buffer (63,192 bytes, version
0x0B=11, rowCount 50 = 25 Alarm.Set + 25 Alarm.Clear). Feeding that real buffer through
HistorianEventRowProtocol.Parse exposed a latent parser bug.
The real buffer layout is: version(2) + rowCount(4) + headerField(4, =0x1E) followed by
MARKERLESS rows (rowFormat(2)=7 + filetime(8) + 8x u16 slots + compact-ascii type +
propCount + props). The parser wrongly treated the one-time 0x1E field as a per-row
marker and re-consumed [marker+format] for every row, so it decoded only the FIRST row
of any multi-row buffer and stopped. This is not gRPC-specific: the captured WCF v9
buffer has the identical 0900 <rowCount> 1E000000 0700 header, so the shipped WCF event
read had the same latent multi-row truncation.
Fix: read a 10-byte buffer header (skip the 0x1E field once) and parse markerless rows;
accept container version 9 (WCF) and 11 (gRPC), mirroring the interface-version gate that
accepts History 11 and 12.
Verified: the real 50-row buffer now decodes to exactly 50 events, ending cleanly at
end-of-buffer (Parse_RealStockClientCapture_DecodesAllEvents, gated on
HISTORIAN_EVENT_CAPTURE_NDJSON so it skips without the gitignored capture), plus a
synthetic v11 golden test. 328 offline tests pass.
The parse path is now verified against the provided client's real event data on both
transports; the only remaining gap for gRPC events is the server delivering rows to our
connection (the documented retrieval-server-gate).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B6mcaT2PjRFKcogzp9UkfC
Joseph Doherty
8f4a188f78