Skip to content

v2.3.3 — Headless command-line scan and export scope fixes

Latest

Choose a tag to compare

@kristofferR kristofferR released this 18 Nov 09:45

This release adds a command-line scan that runs without a window, and fixes export ignoring the selection and the chosen scope.

Command line

  • iptv-checker scan <playlist> checks a saved playlist (by name or id), an M3U file or folder, or a playlist URL with no window and no display, so it runs from cron or a systemd timer on a headless server. It uses the saved settings, and the run is written to scan history, so the app shows it later (#304, ref #262).
  • Every scan setting has an override flag, for example --timeout 5 or --skip-screenshots, and --preset <name> applies a saved preset. --export <dir> writes the CSV report, a playlist per status and the renamed playlist.
  • The exit code is 0 when the scan completed, 1 when it failed, 2 for invalid arguments and 130 when it was interrupted. A playlist with no channels counts as a failure, and so does a network outage the scan notices.
  • A headless run checks every channel again and keeps no checkpoint. Every status playlist is rewritten on export, empty ones included.
  • On Windows the app has no console, so the output has to be redirected to a file. That path is untested.

Export

  • The export menu starts from the selection each time it opens: "Selected" when more than one row is selected, otherwise "All". It always started on "All" (#303, ref #302).
  • "Export Scan Log (JSON)" follows the export scope. A scoped file holds only those channels and records how many were left out as channels_omitted_by_scope; summary still covers the whole run.
  • The File menu's "Export Filtered M3U/M3U8" exports the filtered list. It used the scope last picked in the export menu, "All" by default. The other File menu exports use the selection-driven default.
  • A File menu export chosen with nothing loaded no longer stays queued and runs once a playlist opens.
  • Previewing one row of a multi-row selection from the context menu keeps the selection. The table still showed the rows as selected, but the stored selection collapsed to that row, so "Selected" exported one channel.

Fixes

  • Scan history ids are unique across launches. They came from a per-launch counter, so runs from different launches shared one id and one results file. Existing entries are unchanged (#304).
  • The scan history index is rewritten under a file lock, because the app and a command-line run can write it at the same time.

Internal

  • Scan orchestration moved from the Tauri command layer into the engine, behind a host interface that the app handle and the headless host both implement (#304).
  • The TypeScript types that cross the IPC boundary are generated from the Rust definitions by ts-rs, and CI fails when the committed bindings are stale (#303).
  • CI runs clippy and the Rust tests on Windows. Debug builds on Windows embed the application manifest through the linker so test executables can start; release builds are unchanged (#303).

Full changelog: v2.3.2...v2.3.3