padosoft-crash-triagelisted

# Crash triage A crash report is somebody else's failure, on a device you do not have, in a build you cannot attach to. The work is not finding a plausible cause — plausible causes are cheap. It is **separating what you know from what you inferred**, and writing the difference down. **The rule: never write a claim stronger than the evidence you actually have. A diagnosis that is right in substance and overstated in wording gets corrected in review, and costs more than the original bug.** --- ## 1. Collect the report before interpreting it Copy out, verbatim: - **Version and build number**, device, operating-system version, timestamp. - **Fatal or not.** An error caught by a boundary and one that escaped everything are different events, and the same underlying fault often appears as both. - **Every custom key** the application publishes — the current route, mount counts, session age, feature flags, whether the user was authenticated. - **The breadcrumbs with their timestamps.** A cycle of crash → retry → crash, and how many in how long, is evidence about determinism that nothing else gives you. - **The full stack, with line *and* column.** **An issue with many variants across several builds is a bucket, not a defect.** Reporters group by a signature that is coarser than the cause. Pick one event and work on that one; a "fix" for a bucket is a fix for whichever variant you happened to read. ## 2. Does the build already contain the fix? ```bash git merge-base --is