EQ2Parser 0.5.6

v0.5.6

v0.5.6 - foreign log holders are no longer named in uploads

EQ2Parser 0.5.5

v0.5.5

fix(raid): fight allies are CLASSIFIED before feeding attendance — pe…

EQ2Parser 0.5.4

fix(raid): attendance only starts from a PROVEN roster-macro pair

Live 2026-09-10 disaster: guildies running SIX-MAN content all day
minted dozens of overlapping phantom attendance "raids" with ~36h
durations and the whole guild marked present. Three causes, all fixed:

- A lone whoraid block seeded the raid immediately — but pressed
  outside a raid it echoes "Not in a raid" and completes EMPTY, and the
  guild who then paired against it. whoraid is now only ever the held
  half of the pair, and the pair validates BOTH ways: the raid half
  must be non-empty and the plain half must be a PURE guild block —
  every row detailed (no anonymous) and every row tagged with the SAME
  guild (verified against live guild-who captures: guildmates always
  show full detail to each other). "/who", "/who all", zone whos and
  out-of-raid macro presses can never satisfy this.

- Fight allies and raid deltas fed the roster with no gating — a
  6-man's allies became "the raid". The tracker is now ARMED only by a
  validated pair; until then every passive signal is discarded, and
  TickAttendance additionally refuses to upload unarmed state.

- No session lifecycle: trackers accumulated members for days, so
  first_seen stretched back to the ingest clock clamp (the 2147-minute
  sessions). A validated pair arriving >6h after the previous one now
  auto-rolls a fresh session — Tuesday's roster can't haunt Thursday
  even if nobody presses "New session".

Test corpus extended with the verbatim 6-man capture (whoraid "Not in a
raid" + valid guild who) plus mixed-guild, anonymous-row, unarmed-signal
and session-rollover cases.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

EQ2Parser 0.5.3

feat(diagnostics): session journal + WER dumps + startup post-mortem

A 0.5.x user's app dies with NOTHING in the crashes folder — a death the
CLR never saw (native crash, stack overflow, OOM kill, or an antivirus
terminating the process). CrashLog already covers every managed escape
route; this adds the nets for the rest, all best-effort and silent on
failure:

- SessionJournal (%LocalAppData%\EQ2Parser\session.log): startup-stage
  breadcrumbs, source add marks, raid-command-file writes, a mark BEFORE
  every native TTS synthesis call (prime unmanaged suspect), and a
  once-a-minute heartbeat with working-set/GC numbers (OOM creep shows
  as a marching trend). Ends with a clean-exit mark.
- WER LocalDumps self-registration (HKCU, no admin): Windows writes a
  minidump to crashes\dumps on native crashes and stack overflows.
- Startup post-mortem: a previous session.log without the clean-exit
  mark means the last run died silently — its journal is copied into
  crashes\postmortem-<ts>.txt together with any EQ2Parser-related
  Application Error / .NET Runtime / WER event-log entries from the
  last 48h. Empty harvest + no dump = external kill (antivirus).

Verified end-to-end: run + taskkill /F + restart produces the postmortem
with the journal tail and the "(none found — consistent with an external
process kill)" verdict; the WER key lands with DumpType=1, DumpCount=5.
Support flow is now just "send the crashes folder".

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

EQ2Parser 0.5.2

feat(raid): detect duplicate in-game applications of a confirmed step

An announcement echo that matches a recently CONFIRMED step but no
pending one means the game ran that points command AGAIN — a rapid
re-press inside the log-flush lag, or two officers pressing for the
same item. Previously that second application vanished silently; now it
is flagged in the status bar and appended to dkp-ledger.log as
"DUPLICATE APPLICATION" with the exact command, so the officer can
correct the points in game. Confirmed steps are remembered for five
minutes. Loc x4.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

EQ2Parser 0.5.1

v0.5.1

feat(raid): announced one-step DKP/loot presses, item-link comments, …

EQ2Parser 0.5.0

v0.5.0

feat(raid): loot tracking — auto-synced charge file, min bid, per-row…

EQ2Parser 0.4.0

v0.4.0

Raid tab: roster tracking, DKP to mains with press-until-done awardin…

EQ2Parser 0.3.4

v0.3.4

v0.3.4 — fix dropped second-person multi/flurry/AoE autoattack swings

EQ2Parser 0.3.3

release: move the update feed onto the (now-public) main repo

Consolidate distribution: releases publish to VortexUK/EQ2Parser instead
of the separate EQ2Parser-releases repo, and the in-app updater reads the
main repo. Main-repo publish uses the workflow's GITHUB_TOKEN (same-repo,
no PAT).

v0.3.3 is the transitional build: it's ALSO published to the old
releases repo (RELEASES_PAT) so installs still pointing at that feed
(<= v0.3.2) find v0.3.3 and migrate onto the main feed. Once testers are
on v0.3.3+, the transitional step + RELEASES_PAT preflight come out and
future releases go to the main repo only.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>