Module diagnostics 

Structs

BreakClock

Per-kind interval clock: how long since this kind last fired and the configured interval, so the report can show “due in / overdue by”.

EnvFacts

Host windowing-environment facts, gathered from process env vars. Only meaningful on Linux (X11 vs Wayland, compositor); on macOS / Windows the windowing system is implied by the OS. Split from the gathering so the markdown rendering is unit-testable.

LiveReadings

Volatile readings gathered by the command from the OS / plugins: the idle probe, DND state, screen-lock, notification permission, and OS-level autostart. Bundled and passed into runtime_snapshot_section so that function touches only in-memory scheduler state — these probes go through X11 / dbus / plugin FFI that segfaults on the headless CI runner, so they must stay out of the testable path.

MonitorFacts

One monitor’s geometry for the report. Mirrors tauri::Monitor fields we care about, kept plain so format_monitors is testable.

RuntimeSnapshot

Everything the report needs to answer “why isn’t a break happening right now”. Gathered live from the scheduler + windowing system, then rendered by format_runtime_snapshot (kept pure for testing).

Constants

LOG_FILE_NAME 🔒

REPORT_LOG_BYTES 🔒

Functions

break_kind_label 🔒

build_diagnostics_report

collapse_hint_pool 🔒

display_server 🔒

Best-effort name for the display server backing the app, used to triage overlay/transparency issues that only reproduce on one stack.

environment_section 🔒

format_break_clock 🔒

format_environment 🔒

Render the whole environment block from already-gathered facts. Pure: the caller does the env / windowing-system queries and passes the results in, so this formatting is fully unit-testable.

format_monitors 🔒

format_runtime_snapshot 🔒

gather_monitors 🔒

log_file_path 🔒

log_startup_banner

Gather the startup facts and emit the banner at info level. Called once from the Tauri setup hook.

next_due_secs 🔒

Seconds until the next break of a kind: positive = still counting down, zero or negative = overdue (the scheduler will fire as soon as no guard is suppressing it). i64 so overdue is representable.

os_description 🔒

read_tail 🔒

redact_log_tail 🔒

redact_sensitive 🔒

runtime_snapshot_section 🔒

Build the Runtime section from in-memory scheduler state plus the already-gathered LiveReadings. No OS / plugin / AppHandle access, so it’s testable with a bare scheduler.

startup_banner 🔒

Compact one-line environment summary logged once at startup, so the log file (and therefore every diagnostics report’s log tail) always opens with the context a bug report needs — even if the user never generates a report. Pure so the formatting is unit-testable.