ANetBBS Changelog

Versions are internal build numbers. Public releases are tagged
separately. Current release: v1.0b2.33 (July 2026). Full release: August 1 2026.

v1.0a2.166 — Fix ZMODEM first-attempt failures (June 2026)

features/xfer.py: Removed --escape from ZMODEM send_flags. With --escape, sz sends a ZSINIT frame to negotiate extended character escaping; SyncTERM (and many other terminal emulators) responds with ZRINIT instead of ZACK, stalling the handshake and causing sz to fail or retry. Without --escape the handshake is a clean ZRQINIT → ZRINIT → ZFILE sequence and transfers succeed on the first attempt. sz still escapes XON, XOFF, and DLE by default. --binary is kept to prevent newline translation of binary data.

v1.0a2.165 — (superseded — do not use)

v1.0a2.164 — Fix /imsg/directory crash + RSS short-URL + title word-wrap (June 2026)

templates/imsg/directory.html: Fixed 500 crash on /imsg/directory — current_app is not a Jinja2 global in Flask, causing UndefinedError for any admin visiting the page. Replaced with url_for('admin.registry_self'). Also corrected the endpoint name (admin.registry_self_register doesn't exist).
web/rss.py: Added redirect_bp with GET /r/<item_id> — unauthenticated 302 redirect to the original RSS article URL. No new DB model; uses RssItem.id as the short code.
features/bbs_ui.py _rss_view_item: Replaced truncated raw link with a short URL ({web_base}/r/{item.id}) that always fits in 80 columns. Title now word-wraps at 74 cols instead of hard-truncating.
web_app.py: Registered redirect_bp.

v1.0a2.161 — Fix mixed-art \n stripping; 80-col viewer; ASCII status bar (June 2026)

features/anedit.py launch_aneview: Changed \n strip logic — only strips for pure flat block art (block chars present, no cursor-pos sequences). Cursor-pos art (including mixed flat+cursor-pos like ConstructiveChaos) keeps \n so that flat header sections (e.g. the CONSTRUCTION logo) aren't collapsed into a single overflowing row. Cursor-pos absolute positioning is unaffected by artifact \n between sequences.
features/anedit.py _ViewerScreen: Fixed all widths from _W=79 to 80 — header bar, status bar, and _ansi_trunc limit. Prevents the rightmost art column from being clipped.
features/anedit.py _ViewerScreen.draw_status: Replaced ↑↓ Unicode arrows (rendered as ?? in SyncTERM) with PgUp/PgDn ASCII text.
web/render_msg.py: Same condition change as terminal — strip \n only for flat block art without cursor-pos. Fixes ConstructiveChaos logo being missing in web view.

v1.0a2.160 — VT renderer for terminal; flat-art \n fix; borderless ANView (June 2026)

features/ansi_html.py: Refactored VT engine into _run_vt(text) (returns cell grid). Added to_ansi_lines(text, width=80) — runs body through VT renderer and returns a list of terminal-ready ANSI-coloured strings. Added _HEX_TO_FG / _HEX_TO_BG reverse-palette maps. _to_html_vt now calls _run_vt internally.
features/anedit.py _ViewerScreen: Removed box-drawing border. Row 1 = dark header bar (subject), rows 2–23 = 22-line content area (up from 17), row 24 = status/hints bar. No side borders or separator lines.
features/anedit.py ANView: Overrides _ensure_visible, _scroll_view, _pgup, _pgdn, _doc_start, _doc_end to use the new 22-line viewport.
features/anedit.py launch_aneview: Now handles CP437 decode + pipe-code conversion + art detection internally. Strips \n for cursor-pos or block-art bodies before calling to_ansi_lines(). All message types (ANSI art, cursor-pos art, plain text) render in the scrollable ANView — the raw full-screen display path is gone.
features/bbs_ui.py: Removed cursor-pos detection and raw display path. All messages route to launch_aneview().
web/render_msg.py: render_msg_body / render_msg_body_rich now also strip \n for flat block-art (CP437 block chars, no cursor-pos) before VT rendering. Fixes horizontal-bar scatter for flat art in web view.
echomail/qwk.py: At import time, if body contains ESC (\x1b), \xe3 record-boundary bytes are stripped (not converted to \n). Real line structure in ANSI art comes from embedded \r\n. Plain-text bodies unchanged (\xe3 → \n).

v1.0a2.159 — Fix QWK art scatter: strip record-boundary \n; pipe colors in terminal (June 2026)

Root cause: QWK 0xE3 record separators are stored as \n in the DB. When the body has cursor-positioning sequences, those \n characters were inserting spurious line breaks between records, moving the cursor to the wrong row before the next \x1b[r;cH could reposition it. Each record's content landed on the wrong visual row, producing the scattered-fragment display.
web/render_msg.py render_msg_body / render_msg_body_rich: Now uses to_html() from features/ansi_html.py (VT renderer) instead of the local streaming _ansi_to_html that silently dropped cursor-pos sequences. When cursor-pos sequences are detected after CSI-fix and pipe-code conversion, \n chars are stripped before calling the VT renderer. Flat-art and plain-text messages continue to use the streaming path.
features/bbs_ui.py read_echo_area: (1) Removed _BLOCK_ART fallback — only messages with detected cursor-positioning sequences trigger full-screen raw display; flat art goes to scrollable ANView. (2) For cursor-pos art: \n stripped from body_unicode before streaming to terminal. (3) Added |NN pipe color → ANSI SGR conversion applied to body_unicode before display (handles Synchronet/Mystic pipe codes in message signatures and art).

v1.0a2.158 — Strip SAUCE record from ANSI art before rendering (June 2026)

echomail/qwk.py _parse_messages_dat: Strip the SAUCE metadata record at QWK import time. At the first \x1a (Ctrl+Z) byte in the decoded body, truncate — everything from \x1a onward is the binary SAUCE header (title, author, group, date, resolution fields) that, when passed to a renderer, appears as random ANSI sequences and scatters content across the display. Existing DB messages are handled at display time.
features/bbs_ui.py read_echo_area: Strip the SAUCE record from the body before CSI fix and CP437 decode. Applies to all messages already in the DB regardless of when they were imported.
features/ansi_html.py to_html: Strip the SAUCE record for string input. The bytes-input path already handled \x1aSAUCE stripping; this adds the same protection for the common string path used when rendering messages from the DB.

v1.0a2.157 — ANSI art full-width VT renderer + block-char detection (June 2026)

features/ansi_html.py _to_html_vt: VT renderer now renders every row to the full WIDTH (80 columns) instead of stopping at that row's max_c. Previously, rows with sparse content (art that writes only a few characters at specific columns) appeared as narrow bars in the browser; now all rows are consistently 80 chars wide, preserving the 2-D portrait shape.
features/ansi_html.py to_html: Added _HAS_BLOCK_ART detection. When the body contains CP437 block/half-block characters (█ ▄ ▀ ░ ▒ ▓ or latin-1 mojibake equivalents U+00B0-B2, U+00DB-DF) together with any ANSI escape, the VT renderer is used instead of the streaming renderer. Covers flat ANSI art (no cursor-positioning) that previously got the fast streaming path.
features/bbs_ui.py read_echo_area: (1) Added f to _HAS_CPOS regex (HP-style cursor-pos alias for H, was missing). (2) Added block-char detection as a fallback: CP437 block chars in latin-1 mojibake form trigger raw terminal display even when no cursor-pos sequences are detected. (3) Raw terminal output now converts \n to \r\n so flat-art rows return to column 0 on each new line.

v1.0a2.156 — ANSI art terminal raw display; web CSS fix; ANView colors (June 2026)

features/bbs_ui.py read_echo_area: Detect cursor-positioning sequences in the decoded body. If found (ANSI art), display raw to the terminal (\x1b[2J → stream body → \x1b[25;1H prompt) so full 2-D positioning and colors render exactly as the sender intended — matches Synchronet/SyncTERM behavior. Plain-text messages (no cursor positioning) still use the ANView scrollable frame.
features/anedit.py launch_aneview: Change ANSI strip from removing ALL escape sequences to removing only cursor-positioning sequences ([H f A B C D G J K s u r]). SGR color/attribute codes (\x1b[...m) are now preserved so plain-text colored messages display with colors inside the ANView frame.
features/anedit.py _ViewerScreen.draw_text: New ANSI-aware override. Uses _ansi_trunc() helper to truncate lines at _TW visible characters (ignoring escape sequences) and \x1b[K to erase trailing space rather than .ljust(). Prevents SGR sequences from being sliced mid-sequence and eliminates visible-width miscounts from ANSI bytes.
features/anedit.py _ansi_trunc: New helper — truncates a string with embedded ANSI sequences to a maximum of maxlen visible characters while preserving all escape sequences.
templates/echomail/read.html: Fix CSS for ANSI body display. Changed white-space: pre-wrap (allowed long art rows to wrap mid-row, scrambling the image) to white-space: pre with overflow-x: auto. Removed the per-span white-space: pre-wrap override. Fixed line-height: 1.0 (was 1.5/1.2, adding unwanted gaps between art rows).

v1.0a2.155 — CP437 decode in ANView; VT ANSI renderer for web; scroll fix (June 2026)

features/anedit.py launch_aneview: Body was displayed as latin-1 mojibake (Û instead of █). Apply the same CP437 re-decode as the web path (encode latin-1 → bytes → decode cp437, keeping control chars 0x01–0x1F as-is) before stripping ANSI sequences. Block/line-drawing chars now render as proper Unicode in UTF-8 terminal sessions.
features/anedit.py ANView._handle + _scroll_view: UP/DOWN arrows now scroll the viewport by 1 line immediately (was: moved cy without visible text change until cy left viewport). Space scrolls down a page. PgUp/PgDn/Home/End/Ctrl+Home/End all work. Replaced non-ASCII ↑↓ hint chars with ASCII Up/Dn — SyncTERM displayed them as ??.
features/ansi_html.py: Added virtual-terminal renderer _to_html_vt. When cursor-positioning sequences (\x1b[H, \x1b[A/B/C/D, \x1b[2J, etc.) are detected in the body, routes to the VT renderer which maintains an 80-column cell grid, processes all cursor moves, and outputs the correct 2-D layout. Streaming renderer kept for plain-text messages. Result: ANSI art on the web viewer displays with correct positioning instead of as a scrambled linear stream.

v1.0a2.154 — Terminal ANSI fix, no-pager reader, ANView message viewer (June 2026)

features/bbs_ui.py: Three improvements to the terminal echomail reader. (1) Apply the same CSI-split fix from v1.0a2.153 at view time so QWK 0xE3 artifacts ([1;45m etc.) don't appear in the terminal. (2) Remove the 18-line page-break (-- more (Enter, Q=back to list) --) — messages scroll freely. (3) Add R=Reply and N=New Msg options after reading; both launch ANEdit with the appropriate pre-fills in the current area.
features/anedit.py: Added _ViewerScreen, ANView, and launch_aneview. ANView is a read-only ANEdit-frame viewer: navigation keys scroll, R/N/Q exit with action codes, bottom border shows viewer hints, status bar shows Read-only · Ln:X/Y, cursor is hidden.

v1.0a2.153 — Fix ANSI sequences split anywhere by QWK line separator (June 2026)

echomail/qwk.py _parse_messages_dat + web/render_msg.py render_msg_body/render_msg_body_rich: v1.0a2.152 only fixed the case where QWK 0xE3 fell between ESC and [. The separator can fall anywhere in a CSI sequence — in the parameter string too (e.g. \x1b[1;\n37m). Replaced the narrow \x1b\n→\x1b string replace with re.sub(r'\x1b\n?\[[0-9;?\n]*[@-~]', lambda m: m.group(0).replace('\n',''), ...) which strips \n from any position within a matched escape sequence, covering all split positions in one pass.

v1.0a2.152 — Fix ANSI sequences split by QWK line separator (June 2026)

echomail/qwk.py _parse_messages_dat: QWK byte 0xE3 (line separator) can fall between ESC (0x1B) and [ when an ANSI CSI sequence straddles a QWK line boundary. After \xe3→\n conversion the sequence became \x1b\n[1;45m, which _ansi_to_html's regex couldn't match. Added .replace('\x1b\n', '\x1b') after the separator conversion to rejoin split sequences before storage.
web/render_msg.py render_msg_body / render_msg_body_rich: same \x1b\n→\x1b replacement at render time so all messages already in the database (imported before this fix) render correctly without a migration.

v1.0a2.151 — Fix ANSI rendering in web echomail viewer (June 2026)

web/render_msg.py _decode_charset: CP437 maps byte 0x1B (ESC) to U+2190 (←), breaking _ansi_to_html's \x1b\[ regex so all ANSI color sequences leaked as visible text. Fix: after CP437 decode, restore bytes 0x01–0x1F to their original control-char code points. CP437 is a single-byte codec so a zip(raw, decoded) rebuild is safe and exact.

v1.0a2.150 — QWK overhaul: nine bug fixes (June 2026)

Full audit and repair of the QWK subsystem.

poller.py: stamp _qwk_conf_num on outbound messages before QWKClient.poll(). Without it, all outbound echomail went to conference 0 (netmail) — hub discarded it.
qwk_user.py upload(): REP body decode now uses latin-1 + replace('\xe3', '\n') — CP437 decode turned 0xE3 into π instead of a newline.
qwk_user.py upload(): added network_id=area.network_id to EchomailMessage — NOT NULL column was unset; IntegrityError was silently swallowed and all REP replies lost.
qwk_user.py upload(): moved EchoArea.query outside the parse loop (was running one DB query per message).
qwk_user.py _build_qwk_blob: fixed body line separator \r\n → \xe3; encoding cp437 → latin-1. QWK readers saw blank bodies.
qwk_user.py _last_read_map: r.last_read_id → r.last_message_id. AttributeError was swallowed; every download included all messages ever.
qwk.py _ftp_download: hub_id candidates now listed before packet_id — for DOVE-Net you download VERT.qwk, not ANET.qwk.
qwk.py _parse_messages_dat: skip killed messages (active_flag == 0xE2) after reading body bytes; also added defensive \r\n/\r normalization.
qwk.py _clean_body: extract @CHRS:/@CHARSET: value and return as chrs key — previously discarded, so UTF-8 messages from Synchronet displayed as garbled CP437. _import_message already accepted msg_data.get('chrs'), it just never received it.