The share outside the known companies is now split by site names taken from
the domains of recent queries (cdn.example.co.uk shows as Example). Domains
are read only from the in-memory log; the saved statistics still keep numbers
only.
Queries for the same domain (A, AAAA, HTTPS and other types) now share one
row that shows all requested types, the number of queries and the status
and time of the newest one. Refs #4
New look based on the SparkleDNS concept: a dark theme, a floating
bottom panel with Home, Activity and Settings, and pages for the
blocklists, apps and About.
- Home: big ON/OFF switch, the current server, 24-hour stats and chart
- Activity: requests, blocked, average time and cache share for 24 hours
or 7 days, company dominance, and the live log with search and filter
- Blocklists: HaGeZi Pro preset next to Pro++ and TIF, lists can be
switched off without deleting them, automatic updates (daily or
weekly, optionally Wi-Fi only) through JobScheduler
- Settings: server list, custom server, server test and per-app DNS
- Hourly numbers for the charts are kept on the device, without domains
The server list now ships with the app; it is no longer downloaded
from the repository.
A quick double tap could start the connection twice, open the VPN
dialog twice, or turn the VPN back on right after switching it off.
The buttons are disabled until the VPN state changes (at most 5 s).
Icons were always drawn at 96 px: blurry on screens where 40 dp is
larger, wasted memory where it is smaller. The cache is now limited to
10 MB instead of 256 entries.
The app is now compiled the way Macrobenchmark's Partial mode does it,
so runs compare baseline profiles instead of what the JIT learned in a
warm-up run. Tapping a row scrolls the list when the row is off screen.
contentType for lazy list items, a set for the own-domain lookup per
Monitor row, the group chevron animated in the draw phase, the custom
URL check remembered and the number format reused.
With a slow server all 16 workers can be busy for up to 8 s each, and
queries kept piling up long after the asking app had retried or given
up, delaying every newer one.
Adding or removing one own domain re-sorted the whole combined set,
allocating 8 bytes per entry of every list each time. Lists are now
merged in linear time and own domains are looked up in their own array.
Home and Monitor polled the query log every second and Home queried
ConnectivityManager every 3 s, and kept doing so while the activity
was stopped. The log now signals changes (still at most once a second),
Private DNS comes from a network callback, and all flows are collected
with collectAsStateWithLifecycle.