Skip to content

v1.0.3

Choose a tag to compare

View all tags
@github-actions github-actions released this 20 Jul 19:39
v1.0.3
51c4486

Added: Recovery, concurrency, resource-pressure, and decoder-robustness hardening.

Recovery & lifecycle (reconnect_test.go)

  • TestClientReconnect_AfterServerRestart — reconnect after server shutdown; new-port server model.
  • TestClientReconnect_MultipleSequential — 5 sequential connect/close cycles without resource leak.
  • TestClientReconnect_URCBResubscription — URCB re-established after reconnect on same server.
  • TestClient_NoGoroutineLeakOnConnectDisconnect — 10 cycles; goroutine delta ≤ 3.
  • TestServer_Close_StopsReportEngine — post-close connections refused immediately.
  • TestClientReconnect_ReadAfterServerClose — reads return errors (not hang) after server gone.

BRCB EntryID resume (interop/brcb_resume_test.go, report_engine.go)

  • Implemented decodeEntryID and filterEntriesAfter helpers.
  • enableRCB reads the client's pre-written EntryID and filters buffer replay to entries strictly after that point.
  • Policy: exclusive boundary — ID N means replay entries > N; zero means full replay; future ID means no replay.
  • EntryIDs are 8-byte big-endian, monotonically increasing from 1, non-persistent across server restarts.
  • PurgeBuf=true invalidates all previously issued EntryIDs.
  • Three interop tests: Resume, ZeroResume (full buffer), ResumeFromFuture (no replay).

Bug fix: BRCB replay skipped after unconditional re-enable (report_engine.go)

  • Root cause: when a BRCB generated a buffered entry it wrote EntryID to the value store. On re-enable, enableRCB read that server-written value as the resume filter, excluding the only buffered entry — so no replay was delivered.
  • Fix: EntryID in the store is reset to all-zero bytes when RptEna=false is written. A subsequent re-enable with no client-written EntryID decodes to resumeID=0, which means full replay. Tests that explicitly write an EntryID before re-enabling are unaffected — they overwrite the cleared value before RptEna=true.
  • Covered by TestGoServer_BRCB_Replay (was timing out, now passes).

Concurrent request hardening (concurrent_test.go)

  • Context cancellation releases all in-flight confirmed requests.
  • Asynchronous report delivery does not stall or corrupt concurrent read responses.
  • Aborting the client while 20 writes are in flight — all goroutines exit cleanly.
  • Timed-out requests do not accumulate goroutines (delta ≤ 3 after 10 timeouts).

Resource pressure tests (resource_exhaust_test.go)

  • 50 sequential connections and 20 concurrent connections without descriptor leak.
  • Report flood (30 GIs, queue-size=2) — reads succeed throughout; OverflowDropNewest drops silently.
  • 10×200 concurrent SetValue with 100 simultaneous client reads — zero read errors.

SCL corpus expansion (scl_corpus_test.go)

  • Multi-LN model (LLN0 + 3×GGIO + MMXU with AnalogueValue DAType/BDA).
  • Multi-LD model (4 logical devices).
  • SDO chain (DPC with Oper sub-object).
  • Enum DAs (ENS with custom enum type).
  • Multi-LN URCB report with cross-LN dataset members.

Report decoder robustness (additional fuzz targets in fuzz_test.go)

  • FuzzDecodeReportIndication — arbitrary value arrays (4 seeds: minimal, empty, all-zeros, all-ones).
  • FuzzDecodeOptFlds — arbitrary bit-string bytes for the OptFlds decoder.
  • FuzzDecodeTrgOps — arbitrary bit-string bytes for the TrgOps decoder.
  • FuzzDecodeReportIndication_NilValues — selective nil entries via fuzz-driven bit masks.

API documentation (example_test.go)

  • Added Example* functions for Dial, SubscribeReport, Write, GetReportControlBlock, NewServer, SetValue, RegisterControl, and ParseRef.

Service profile (INTEROP.md)

  • IEC 61850-7-2 service table (client + server columns).
  • CDC conformance table with interop-tested column.
  • Full report features table (URCB/BRCB, all OptFlds).
  • New BRCB EntryID resume section with explicit exclusive-boundary contract, byte format, stability guarantees, overflow interaction, and purge invalidation.
  • New Report backpressure and loss policy section documenting URCB vs. BRCB queue loss semantics.
  • Known limitations and deferred features table.

Scheduled fuzz workflow (.github/workflows/fuzz.yml)

  • Nightly 5-minute fuzz run per target, with workflow_dispatch override.
  • Failing corpus entries uploaded as artifacts (30-day retention).

Race detector: go test -race ./... passes cleanly in both go-iec61850 and go-mms.

Assets 18
Loading