Journal
Notes from the traces, not from a transformation deck.
Short pieces on crash evidence we actually sit with: One UI ANRs, tombstones, mapping files, Play’s bad-behaviour door, and WebView clusters after a Chrome train.
11 June 2026
Why One UI ANRs show up in Malaysia before they show up on your office Samsung
A flagship One UI unit in the office often has more RAM and a calmer battery governor than the A-series handset your grocery app actually runs on in Kepong and Kajang.
2 May 2026
Reading an Android tombstone without drowning in the unwind
Native crashes arrive as tombstones. The useful part is smaller than the file looks: abort message, fault address, and the first frames that belong to your .so files.
17 April 2026
The bad-behaviour threshold is a door, not a score
Play’s user-perceived crash threshold is a gate on discoverability. Treating it as a vanity score leads teams to “improve the percentage” without naming a single stack.
8 March 2026
Mapping files: why Crashlytics says unknown after a perfectly ordinary R8 run
If mapping.txt for that version code never left CI, every obfuscated Java frame in the export is honest noise. The audit cannot invent the original names.
21 January 2026
When a Chrome update turns into your WebView crash cluster
A Monday cluster that names libwebview or Chrome in the tombstone is often a vendor WebView update, not a feature you shipped on Friday.