rdmsm4x-changelog-20260916-1234-replicantdb-sigpipe-issue-20260916-14

Remediated pre-existing SIGPIPE crash in replicantDB ProcessFleetSearchTransport that caused swift test to exit rc=1 at test #522 mid-suite.

• Host: rdmsm4x
• Scope: replicantDB (Sources/ReplicantDBCore/FleetSearch.swift, Tests/ReplicantDBTests/FleetSearchTests.swift)
• Ticket: ISSUE-20260916-14 (Resolved)
• Worktree: /Users/richh/dev/_worktrees/replicantdb-sigpipe-issue-20260916-14
• Branch: fix/replicantdb-sigpipe-issue-20260916-14 (commits 56bce01, 5dc08dc) pushed to backup and fleet remotes.

Root Cause

In ProcessFleetSearchTransport.run, the parent process writes script to stdinPipe.fileHandleForWriting. When the child process (/bin/sh -s or /usr/bin/ssh) terminates early without reading all bytes from stdin (e.g. matching CLI schema marker early, connection failure, or syntax/flag error), writing to the pipe whose reader has closed delivered POSIX SIGPIPE to the caller. Under default signal handling (SIG_DFL), this abruptly killed the xctest process mid-run with signal code 13, masking the true test outcomes.

Remediation

• In Sources/ReplicantDBCore/FleetSearch.swift, wrapped stdinHandle.write in a localized signal(SIGPIPE, SIG_IGN) block with defer { signal(SIGPIPE, previousSignal) }. POSIX EPIPE / Cocoa error 512 is safely caught and discarded without killing the host process or leaking signal modifications.
• In Tests/ReplicantDBTests/FleetSearchTests.swift, added regression test testProcessTransportDoesNotCrashWithSIGPIPEWhenChildExitsWithoutReadingStdin validating SIG_DFL resilience against early-exiting child processes.
• Recorded root cause analysis and test metrics in SESSION-STATE.md.

Verification Evidence

• Negative Control (before fix / reverted fix):
  • Command: swift test --filter testProcessTransportDoesNotCrashWithSIGPIPEWhenChildExitsWithoutReadingStdin
  • Output: error: Process '.../xctest ...' exited with unexpected signal code 13
  • Return Code: 1
• Full Suite After Fix:
  • Command: swift test
  • Result: 1,346 passed, 0 failed, 0 unexpected, 0 crashes (234.141 seconds)
  • Return Code: 0
• Ticket & Mail:
  • Appended comment and resolved ISSUE-20260916-14.
  • Broadcasted resolution on agent_msg.zsh and synced across fleet.