Bug Reporting Pipeline (bonedigger-report)#

Overview#

The bug reporting pipeline is a privacy-first, on-device diagnostic collection system built into Bluefin. Users run ujust report — a shell script (bonedigger-report) collects system state, scrubs PII locally, lets the user review everything, then uploads a GitHub Gist and optionally files a bug. Nothing leaves the machine until the user explicitly confirms it.

Entry point: /usr/libexec/bonedigger-report — invoked by the ujust report recipe in projectbluefin/common.

The pipeline integrates with the bonedigger lifecycle bot, which manages the downstream issue flow: filed → approved → queued → claimed → done 1.

User Commands#

ujust report # collect diagnostics, review, upload to gist, file issue
ujust confirm <issue#> # record that you also hit an existing bug (priority signal)
ujust verify <issue#> # confirm a shipped fix works; closes the loop

confirm count is used by the bonedigger bot as a priority escalation signal 1.

What Gets Collected#

All data is gathered from the local machine only. Source fields 2:

The boot end-state classifier produces exactly four buckets: clean shutdown, suspend-entered-no-resume, abrupt-end-after-resume, or abrupt-end-no-markers 3. The Linux version kernel line is intentionally excluded from the hardware fingerprint to avoid leaking build host strings 4.

PII Scrubbing#

All scrubbing is on-device, before upload. Two scrubber functions are applied layered 5:

• scrub_kernel_log() — MAC addresses → [MAC-REDACTED], IPv4/IPv6 → [IP-REDACTED], UUIDs → [UUID-REDACTED], disk serials (eui/naa/wwn) → [SERIAL-REDACTED], home paths → [REDACTED]. MAC runs before IPv6 to prevent mislabeling.
• scrub_journal_log() — wraps scrub_kernel_log and additionally strips USER=, LOGNAME= env vars and email addresses.

Additional scrubbing applied inline 6:

Optional Deep Hardware Metrics (OpenTelemetry)#

Gated on the presence of /usr/share/ublue-os/otel/ujust-report-config.yaml in the image 7. When the user opts in:

1. otelcol-contrib runs for 35 seconds, writing two spec-compliant OTLP JSON Lines files 8:
   • metrics.otlp.jsonl — CPU, memory, disk, filesystem, network, paging, processes, Podman containers
   • logs.otlp.jsonl — journald service errors/warnings + kernel dmesg
2. Binary resolution order: $HOME/.local/bin/otelcol-contrib → /usr/local/bin/otelcol-contrib → command -v → Podman fallback using docker.io/otel/opentelemetry-collector-contrib 9.
3. The Podman fallback mounts /proc, /sys, /var/log/journal, and the Podman socket — it runs privileged with a 45-second timeout 10.

Inspect locally: jq -c '.' metrics.otlp.jsonl

Upload Flow#

The full flow 11:

1. Report rendered via glow + gum pager for local review
2. User confirms upload with gum confirm — cancellation preserves files in $REPORT_DIR
3. If gh is authenticated: gh gist create --public uploads summary.md + journal.txt (always) + OTel files (if captured)
4. If not authenticated: summary.md is copied to clipboard via wl-copy or xclip; issue URL is shown
5. User is offered gum choose: "File a bug report" / "Request a feature" / "Skip"
   • Bug reports route to the image's own tracker (auto-detected from image-info.json)
   • Feature requests always go to projectbluefin/common

Issue routing by image 12:

Report Output Structure#

$XDG_RUNTIME_DIR/ujust-report/report-XXXXXX/
  summary.md — Markdown report (always)
  journal.txt — Current boot system/service logs (always)
  metrics.otlp.jsonl — OTel host/container metrics (if OTel captured)
  logs.otlp.jsonl — OTel journald + kernel logs (if OTel captured)

A persistent copy is saved to $HOME/.local/state/ujust-report/last/ after successful upload 13. The temp directory is cleaned on EXIT trap.

Key Files#

Ownership rule: The recipe and OTel config are image content — they live in projectbluefin/common and ship via common.bst. Do not copy them to other repos or write sync workflows. 15

Dependencies#