Show signal-killed jobs as failed instead of running forever

[mattdawkins]

Problem

A pipeline (or any desktop job) whose process dies from a signal — e.g. SIGSEGV from a native crash, or SIGKILL from the OOM killer — is reported by Node with a null exit code. The desktop job UI treats exitCode === null as still running:

• the Jobs History row shows an indeterminate progress bar and a spinning icon (JobsHistory.vue),
• the running badge stays lit,
• the failure dialog is suppressed.

So a crashed job appears to run forever and the user is never told it died. This was hit in practice by a SAM2 segmentation pipeline that segfaults mid-run — the process exited (Process exited with code null) but the job spun indefinitely.

Fix

Normalize a finished job (endTime set) whose exitCode is still null to a conventional "killed by signal" code (139) at the single updateHistory funnel. Every downstream exitCode === null check then renders it as a failed job, and the failure dialog explains it was terminated by a signal (crash / OOM).

Also key runningJobs off endTime rather than exitCode, so "running" is defined unambiguously as "has not reported an end time."

Test plan

• Run a pipeline whose process crashes (segfault / OOM). The job now shows as Failed (red ✗, stopped progress bar) with a dialog stating it was terminated by a signal, instead of an endless spinner.
• Normal success (exit 0) and user cancellation (SIGTERM → 143) are unaffected.

🤖 Generated with Claude Code