blankres (0.2.4) unstable; urgency=medium

  * Make the per-executable rate limit configurable as rate_limit_per_hour,
    and log suppression at info rather than debug. Six crashes an hour is
    right for a desktop and wrong for anyone testing, and a silently
    suppressed crash is indistinguishable from a broken daemon.

 -- Herpiko Dwi Aguno <herpiko@gmail.com>  Tue, 08 Sep 2026 00:55:45 +0700

blankres (0.2.3) unstable; urgency=medium

  * Accept reports without a token when the server has none configured, so an
    open endpoint is a deliberate deployment rather than a server that
    rejects everything. The mode is stated at startup, with a warning when it
    is open.
  * Serve a dashboard at / listing the 50 most recent crashes, styled after
    the IRGSH chief dashboard and refreshed with htmx. Report fields come
    from clients and are escaped.

 -- Herpiko Dwi Aguno <herpiko@gmail.com>  Tue, 08 Sep 2026 00:45:16 +0700

blankres (0.2.2) unstable; urgency=medium

  * Log one line per request in the ingest server. It previously logged
    nothing after startup, which is indistinguishable from doing nothing.
  * Include the binary's own crate in the default log filter. The binary is
    blankres-ingest, so its messages are tagged blankres_ingest rather than
    blankres_server, and a filter naming only the library silenced every line
    it logged, startup included.
  * Allow BLANKRES_PUBLIC_URL, BLANKRES_MAX_PAYLOAD_BYTES and
    BLANKRES_UPLOAD_TOKEN_TTL_SECS to be set from the environment, for
    deployment behind a reverse proxy.

 -- Herpiko Dwi Aguno <herpiko@gmail.com>  Tue, 08 Sep 2026 00:22:12 +0700

blankres (0.2.1) unstable; urgency=medium

  * Treat closing the crash window as "Don't send". Dismissing a consent
    dialog is not consent, and leaving the report behind also stopped the
    window from ever opening again, because the path unit that starts it only
    fires when the crash directory goes from empty to non-empty.
  * Include the crash signature in the report filename. Two different bugs in
    one program shared a filename, so the second crash silently overwrote the
    first and took its upload token with it.
  * Also watch the crash directory for changes, so a crash that arrives while
    another report is still pending opens the window.

 -- Herpiko Dwi Aguno <herpiko@gmail.com>  Mon, 07 Sep 2026 21:49:27 +0700

blankres (0.2.0) unstable; urgency=medium

  * Show a crash to the user even when the crash server cannot be reached.
    The daemon previously spooled the report and returned silently, so a
    crash that happened while the server was down was never surfaced at all.
  * Ask the server whether it wants a payload while preparing the local half
    of one, rather than one after the other, so the window never waits on the
    network. The request now has a short deadline, configurable as
    directive_timeout_secs.
  * Default to https://kres.blankonlinux.id as the crash server.
  * Make "Don't send" the default action and place it rightmost. Uploading a
    snapshot of one's own process memory should be deliberate.
  * Translate the interface into Bahasa Indonesia, selected from the locale.
  * Stop the window handing its activation to an already running instance,
    which made it show a stale crash directory or nothing at all.
  * Stop tracking build output in the source tree, which broke source-only
    builds elsewhere.

 -- Herpiko Dwi Aguno <herpiko@gmail.com>  Mon, 07 Sep 2026 21:16:34 +0700

blankres (0.1.0) unstable; urgency=medium

  * Initial release.

 -- Herpiko Dwi Aguno <herpiko@gmail.com>  Mon, 07 Sep 2026 03:34:00 +0700
