xlqy_new2007 — 新仙侣情缘之飘渺纪元 (New Immortal Companion's Fated Love: Ethereal Era)

Archive: xlqy_new2007.rar. Port: 40022. Status: done (boots clean, connects, plays through GB/BIG5 selection and into the age-verification gate with zero fatal errors).

What this is

"ES II mudlib" lineage again (adm/obj/master.c's header comment: "for ES II mudlib... rewritten by Annihilator"), same family as rzrmud (#23) and xkx2001 (#25). Archive extracts to a 复件 xlqy_new/ ("copy of xlqy_new") folder — note the leading Chinese text + space in the directory name, handled fine by convert_lib.sh/cp -r as long as it's quoted throughout.

Found and fixed a real bug in extract.sh while processing this archive: the .rar extraction branch cds into the destination before referencing the (relative) archive path, so unrar couldn't find it and silently produced an empty raw/ while the script still reported success (no -e, exit 0 either way). Fixed in scripts/extract.sh — ARCHIVE is now resolved to an absolute path up front, and the script now checks raw/ actually has files and fails loudly if not, for any archive type. See AGENTS.md's "Archive tooling" section for the writeup; this could in principle have silently affected other .rar archives run through the old script, though every previously-committed lib's work/ already passed its own boot test, so nothing already marked done is suspect.

Fixes applied

1. extract.sh (see above — a tooling fix, not lib-specific). 2. adm/daemons/convertd.lpc:258: the exact AGENTS.md §8h stray- trailing-backslash-before-closing-quote typo in the Greek-alphabet conversion table ("Ρ\", → "Ρ", — this lib's specific character was Rho, not Alpha like the previously-seen instances), recurring for the third time in this "西游记/ES"-adjacent family (xyj2000f, mhxy, now here). This file has plain LF line endings (not CRLF like mhxy's copy) — used sed -i -E 's/\\"(,)?$/"\1/' directly; caught and had to correct a mistake: initially reused the CRLF-safe pattern from AGENTS.md §8h verbatim (s/\\"(,)?\r?$/"\1\r/), which unconditionally appends \r in the replacement even when the original line had none, injecting a spurious \r into this LF-only file's one fixed line. Verified with cat -A before and after; stripped the stray \r back out. Lesson for next time this recurs: check whether the file actually has CRLF (file <path> or cat -A a sample line) BEFORE choosing which of the two sed patterns to use — don't default to the CRLF-safe one "just in case", since its replacement isn't actually conditional on the match. convertd.lpc is #included into many other daemon files, so this one fix should improve the lpcc pass rate substantially (initial sweep pre-fix: 8845/9060 = 97.6%, with hundreds of the failures being "In file included from .../convertd.lpc" cascades — re-swept after the fix, see below).

Observed but not diagnosed: one non-fatal runtime error during login

The connect-time banner sequence prints the configured default error message string ("你发现事情不大对了，但是又说不上来。") once, between the Beijing-time line (chinesed.lpc's bj_time()) and the registered- player-count line — i.e. some error is happening around logind.lpc's UPTIME_CMD->report() call, but it's evidently caught/ non-fatal (execution continues normally to the age-verification prompt and beyond) and produced no trace at all in debug.log (same silent-error shape as AGENTS.md §15d, though not chased down with the same catch()-instrumentation technique here since it isn't blocking anything). Flagging for whoever revisits this lib's cmds/usr/uptime.lpc or UPTIME_CMD->report() next.

Interactive test result

Boots clean, connects; ASCII-art banner + Chinese render correctly, GB/BIG5 selection prompt works, uptime/registration-count lines print (with the one non-fatal error noted above), reaches the age-verification ("are you a student?") gate with zero fatal errors. Did not proceed further into character creation (out of scope for this pass).

lpcc sweep

Initial sweep (before the convertd.lpc fix): 9060 files, 8845 pass / 215 fail (97.6%), with a large fraction of the failures being convertd.lpc inclusion cascades. Re-run after the fix — see the fresh numbers in this lib's lpcc_fail.log (not hand-copied here to avoid this note going stale relative to the regenerable file itself).

Post-hoc fix: UTF8-native is_chinese/registration (AGENTS.md §15h)

Applied in a later batch pass across the whole project: is_chinese/is_chinese2 in the shared chinese.lpc simul_efun fragment used GBK byte-range checks that silently never match real Chinese text once strings are UTF-8 (this driver's str[i] returns a Unicode codepoint, not a GBK byte). This broke character registration specifically -- any real Chinese name was rejected. Fixed the range check to test the CJK Unicode block instead, and halved the GBK-byte-calibrated length bounds in check_legal_name to match. See AGENTS.md §15h for the full writeup; confirmed via a real interactive registration test (Chinese surname + given name reaching the next prompt).

2026-07-23: driver rebuild retest + LPC formatter + WASM check

• Formatter: ran tools/lpc-syntax's format-corpus.mjs over all 9060 .lpc files in work/; 9043 written, 6 already-conformant, 11 errors (files the formatter refused to touch — expected/fine on messy legacy code, not chased individually).
• Native retest: rebuilt ~/src/fluffos/build-debug/src/driver booted this lib clean (zero fatal errors in log/debug.log). Went further than the previous pass's testing (which stopped at the age-verification gate) into a full end-to-end registration: real Chinese name 秦淑, gb/no/new/english-id/Chinese-name/password/email/ gender/gift-allocation (9+y confirm) all completed, landed in 聚见亭; look showed the correct room, score produced a correct full character sheet, quit gave the correct farewell text. Zero debug.log errors this session — the previously-noted "one non-fatal unexplained runtime error during login" from the original pass did not reproduce here.
• WASM: booted cleanly under build-wasm (only expected non-fatal preload warnings). Full registration flow (real Chinese name 秦岚) completed successfully end-to-end under WASM too, reaching 聚见亭; look/quit worked correctly. Same cosmetic query_ip_number()-under-WASM glitch as its xlqy_early sibling (last-connected-address line shows a garbled/timestamp-like value instead of an IP) — non-fatal, doesn't gate login. Verdict: fully playable under WASM.

WASM-enablement pass (loopback-allow / gate bypass / admin seed)

Standard WASM-first pass per AGENTS.md §1.3b/e and §1.5. Gates patched (same shapes as sibling xlqy_early):

• adm/daemons/band.lpc — added is_loopback_site(string) helper (loopback / empty / malformed-IP → local) and short-circuit return 0; at the top of is_banned(), create_char_banned(), is_strict_banned().
• adm/daemons/logind.lpc — added is_loopback_conn(object) helper and:

- logon() reconnect-flood throttle (last_ip==ip && time+10>time, ~line 96) — loopback exempt. - logon() per-IP connection cap (logon_cnt > 10, ~line 113) — loopback exempt. - encoding() !ip_name destruct + dotted-quad char-loop validator (~lines 195-207) — wrapped in if (!is_loopback_conn(ob)); this was the WASM-fatal gate. - No uptime() startup-grace destruct gate; the uptime() at ~line 155 is the 300s newid registration window (content) — left intact.

Admin account seeded: id fluffos, pw Mud@2026, name 浮浮, granted (admin) via fluffos (admin) line prepended to adm/etc/wizlist. Verified: real registration flow (gb/no/new/fluffos/浮浮/pw×2/email/m/ 9/y → entered world), then relogin as fluffos → status (admin), update /adm/daemons/logind → 成功. Retest: fresh normal registration (秦风, gender f) end-to-end with look/score/quit all correct; test char save removed; zero new debug.log errors. Save file: work/data/user/f/fluffos.o (NOT gitignored — orchestrator adds normally, no force-add needed).

Fail-closed retrofit for the loopback-allow gate (2026-07-24)

The loopback helper above was originally written matching the project-wide convention at patch time (AGENTS.md §1.3b), which also treated an empty/non-string/malformed query_ip_number() result as loopback, defensively, because the WASM driver used to return garbage there. That underlying driver bug is now fixed (fluffos commits e33bb5da "fix: query_ip_number() returned uninitialized garbage under WASM" and 007bb863 "feat: synthetic resolve() on WASM instead of raising an LPC error", both 2026-07-23; the locally-built build-debug/build-wasm binaries already postdate both commits), so treating unparseable IPs as trusted is a fail-open gap with no remaining justification. Retrofitted to fail-closed: loopback is now strictly ip == "127.0.0.1" || ip == "::1" || ip[0..3] == "127." (with a stringp() guard before the slice) — a malformed/empty IP now falls through to the NORMAL gate instead of being treated as local. Retested after tightening: fresh driver boot clean, fluffos loopback login and its wizard update command both still work; zero new debug.log errors.

深度功能测试 / Deep functional test (2026-07-24)

First real *playthrough* pass on this lib (all prior passes stopped at registration + look/score/quit; see AGENTS.md §10.7). Native driver (build-debug), one continuous scripts/mudclient.py session per test step (state persists across steps via the lib's own save/relogin path).

Test character: id shenlim, Chinese name 沐阳 (Muyang), password MudPass123, email [email protected], gender male, gift allocation accepted as rolled (体格20/根骨20/悟性25/灵性25). Final state left behind: joined 东方武馆 (Dongfang Martial Hall, via 东方博玉's ask boyu about join), learned 扑击格斗之技 (unarmed, level 1) from teacher NPC 东方聪 via the organic learn unarmed from dongfang cong path, combat_exp 6 (from one real fight, see below), one 粗布衣 (linen robe) in inventory, died once and was successfully reincarnated (气/血 currently 重伤/受伤 from that), sitting at 聚见亭 (START_ROOM) after a final clean quit. Save files: work/data/user/s/shenlim.o, work/data/login/s/shenlim.o — kept intentionally as evidence of a working end-to-end playthrough.

What was tested and confirmed working

• Read help newbie first (work/doc/help/newbie) — correctly described the actual starting flow (南城客栈 is flavor-text/legacy; the real, current onboarding path is 聚见亭 → out → 仙石 in 花果山) and named the real command set (learn X from Y, wimpy, skills, fight/kill).
• Registration: real Chinese name (沐阳), full flow (gb/big5 → age-gate → new → English id → Chinese name → password×2 → email → gender → talent roll 9+confirm y) — reached 聚见亭, look/score both correct.
• Exploration / navigation: read room .lpc source (d/ourhome/xiaoting.lpc, d/dntg/hgs/*.lpc) to route from 聚见亭 through out (→ 仙石/花果山, the actual "leave home, start the journey" transition — confirmed intentional, not a bug, via work/doc/help/start's narrative text) through 后山小路 → 林间小路 → 花果山 → 傲来国 → 北城门 → 北菀街 → 傲来台 → 东苑街 → 东方武馆, ~15 rooms total. Compound diagonal exit names (eastup, southdown, …) initially looked like a rendering bug (这里明显的出口是 eastup、down 和 southdown) but are a confirmed, lib-wide intentional convention (hundreds of legitimate uses across work/d/), not a bug.
• Combat: no dedicated safe-sparring dummy is actually reachable — d/city/obj/muren.lpc and d/obj/misc/muren.lpc (the stat-mirroring accept_fight() training-dummy pattern AGENTS.md §10.7 recommends looking for) exist but are never placed in any room's "objects" map (grepped the whole work/d/ tree for both paths — zero references); orphaned/dead content, not chased further (content gap, not a programming bug). Used help newbie's own recommended fallback instead — the weakest listed mob shape (小猴子/monkey, combat_exp 50, daoxing 0) at 后山小路 — kill xiao houzi produced a full real fight with correct turn-by-turn narration, skill-up message, and (since the character had no armor/skill yet) an actual death, which flowed into a full live test of the death/ reincarnation system (see below) rather than a wimpy-flee.
• Death/reincarnation: feature/damage.lpc's check_gameover() → DEATH_ROOM (/d/death/gate) → d/death/npc/pang.lpc's init()-triggered death_stage() call_out chain (5 staged messages, 5s apart) → reincarnate() → move to REVIVE_ROOM (/d/ourhome/kedian, 荒郊小店) all fired correctly and automatically (no player input needed, matching the design). Confirmed the player's saved startroom field was not touched by any of this (grep -o '"startroom"' work/data/user/s/shenlim.o → no match, both before and after) — no instance of the AGENTS.md §7.24 death-code- overwrites-startroom bug class here.
• Sect/faction join: ask dongfang boyu about join (东方博玉, 馆主/hall master, in 前厅) — age/combat_exp gate checked and passed, wuguan/join flag set, confirmed via the teacher NPC's skills command then unlocking (只有巫师或有师徒关系的人能察看他人的技能 before joining → full skill list after).
• Skill learning, organic teacher path: learn unarmed from dongfang cong in 武馆教练场 — real skill-up message, skills confirms the new skill persisted.
• Shop purchase: buy jitui from xiao er in 荒郊小店 (list showed prices in 文钱/两银子, purchase correctly deducted mixed coin+silver currency) then eat jitui correctly raised the 食物 stat (confirmed via hp) and correctly refused once full (你已经吃太饱了，再也塞不下任何东西了).
• Quit / relogin / state persistence: clean quit → immediate debug.log grep (zero new 错误/error/Fatal/recursion lines) → relogin confirms skill + inventory persisted (location resets to START_ROOM on a clean quit+relogin, since startroom was never explicitly saved by this character — confirmed this is the lib's actual intended behavior, not a bug: cmds/usr/quit.lpc never writes "startroom", and no valid_startroom-flagged room's "make this home" command was ever invoked in this playthrough).
• Net-dead (unclean disconnect) + reconnect: tested both a prompt reconnect (well inside the 15s do_net_dead grace) and a reconnect after a real ~150s wall-clock wait (well past the grace, short of the 600s NET_DEAD_TIMEOUT) — both correctly restored the exact prior room/state via obj/user.lpc's reconnect() (remove_call_outs the pending dumps, prints 重新连线完毕). No instance of the AGENTS.md §7.20/§7.21 void-parking or stuck-wizard-flow classes.
• Full-duration net-dead timeout: happened for real (if accidentally, from research time between test steps rather than a deliberate sleep) partway through this pass and found the two real bugs below — see write-up. Re-verified the fix afterward with a *deliberate*, short-duration repro (NET_DEAD_TIMEOUT temporarily patched from 600 to 20 in include/user.h, reverted immediately after — confirmed clean via git diff showing no residual change).

Bugs found

1. Fixed — adm/simul_efun/message.lpc's tell_room() wrapper passes a raw int 0 as message()'s exclude argument on every 2-arg call. This is AGENTS.md §7.12 ("Shared message/wrapper argument bugs"), recurring on a new lineage — not a new bug class.

• Symptom, found live: debug.log/stdout showed 执行时段错误：*Bad argument 4 to EFUN message() Expected: object, array, Got: int(0)., thrown from obj/user.lpc:236's tell_room(environment(), ...) inside user_dump(DUMP_NET_DEAD) — the exact same file/function AGENTS.md §7.12 already documents from dtsl's deep functional test. Per that section, this abort happens in a call_out-driven context with no enclosing catch(), so it doesn't just log an ugly line — it aborts the rest of user_dump(), at minimum case DUMP_NET_DEAD's follow-up QUIT_CMD->main(...) call right after it (the actual force-quit). Moments later in the same driver session, the whole native process crashed (see bug 2) — matching §7.12's already-documented escalation almost exactly (that section's own dtsl finding: "two characters hitting the aborted user_dump() at nearly the same real moment ... immediately followed by an actual native driver process crash — a C-level double-free abort ... inside dealloc_object()/free_svalue()"). This session had several overlapping net-dead/abandoned-registration connections in quick succession right before the crash (multiple registration attempts — linqing, muyang, shenlin — each left an interactive object net-dead when mudclient.py's connection timed out), so the "multiple simultaneous aborted user_dump()s" precondition plausibly applies here too.
• Root cause: varargs void tell_room(mixed ob, string str, object *exclude) (adm/simul_efun/message.lpc:48) forwards exclude straight into message("tell_room", str, ob, exclude) — an omitted varargs array parameter is the LPC value 0, not ({}), and message() accepts an *omitted* 4th argument but rejects one explicitly passed as the wrong type. Every 2-arg tell_room() call site in the whole lib (717 total call sites of tell_room(, an unknown but nonzero fraction of them 2-arg) was a live, if intermittent, crash risk — not just the two in user_dump().
• Fix (adm/simul_efun/message.lpc:48-56):

``lpc // BEFORE: varargs void tell_room(mixed ob, string str, object *exclude) { if (ob) { message("tell_room", str, ob, exclude); // AFTER: varargs void tell_room(mixed ob, string str, object *exclude) { if (ob) { if (!exclude) exclude = ({}); message("tell_room", str, ob, exclude); ` Fixing the single shared wrapper (per AGENTS.md §7.12's own prescribed remedy) covers all 717 call sites at once, rather than patching individual call sites piecemeal (an earlier draft of this fix instead null-guarded the three tell_room(environment(), ...) call sites in obj/user.lpc directly — reverted once the actual root cause was traced to the shared wrapper, since a per-call-site guard would have left every *other* 2-arg tell_room()` call in the lib equally exposed).

• Verified: reproduced-then-fixed via a deliberate, short repro — temporarily changed include/user.h's NET_DEAD_TIMEOUT from 600 to 20, rebooted, net-deaded shenlim, waited ~45s (past the 15s do_net_dead grace + 20s timeout). Pre-fix this reliably reproduced the exact Bad argument 4 trace; post-fix (this wrapper patched), the same repro produced zero 错误/Bad argument output, the character was cleanly force-quit (confirmed via a normal fresh relogin working immediately after), and the driver process stayed alive. Reverted NET_DEAD_TIMEOUT back to 600 immediately after (confirmed via git diff — no residual change) and did one final clean reboot + full relogin (look/score/skills/i/quit) to confirm the fix in isolation, zero new debug.log errors.

2. Observed, not independently fixable at the LPC level — a genuine native driver crash, matching AGENTS.md §10.8's already-cataloged "driver-fatal ref-count-0 free_svalue() abort" class (third corroborating occurrence, and — per bug 1 above — plausibly the exact mechanism §7.12's own severity-escalation paragraph already predicts, not a fresh mystery).

• What happened: partway through this session (naturally, from an extended real-wall-clock gap between test steps rather than a deliberate wait — see AGENTS.md §10.7 checklist item 8's honesty requirement: this was NOT a scripted repro, it just happened), the driver's own stdout (captured to /tmp/xlqy_new2007_stdout.log, outside the repo) showed:

`` 执行时段错误：*Bad argument 4 to EFUN message() ... 呼叫来自：/obj/user.lpc 的 user_dump() 第 236 行 ... ****** FATAL ERROR: FATAL: Object 0x55c032a76368 /std/skill ref count 0, but not destructed (from free_svalue). (current object was /d/ourhome/kedian2) ... crash() in master called successfully. Aborting. ` The process exited outright — every connection dropped, ss -tlnp showed nothing on port 40022, the driver PID was simply gone. debug.log` showed zero** trace of any of this (same total silence AGENTS.md §10.8 already documents) — the only reason this was caught at all was having the driver's own stdout redirected to a file, per that section's own "actionable takeaway."

• Secondary, compounding finding matching AGENTS.md §7.11 (missing runtime directory): the driver's own crash handler (adm/obj/master.lpc's crash() → log_file("nosave/CRASHES", ...)) itself failed with *Wrong permissions for opening file /log/nosave/CRASHES for append. "No such file or directory" because work/log/nosave/ doesn't exist in this checkout — meaning even the driver's own attempt to leave a permanent crash record for next time silently failed too. Not fixed as a committable change: the entire libs/*/work/log/ tree is gitignored project-wide (.gitignore:12-14), so creating this one directory locally doesn't persist past this session and isn't a real fix at the lib level — flagging for the orchestrator in case a project-wide fix (e.g. scripts/pack_lib_for_web.sh or a boot-prep step always mkdir -p log/nosave/) is worth doing centrally.
• Not fixed / not independently root-caused at the LPC level, same as the two prior §10.8 occurrences (xjcq2000, shiji) — this is a driver-level internal-consistency failure, not a mudlib source pattern. Bug 1's fix removes ONE plausible contributing trigger (the aborted user_dump() mid-execution) but is not proven to be the sole cause; re-verifying bug 1's fix in isolation (see above) did *not* reproduce the FATAL crash in a short (~45s) repro window, which is weakly consistent with bug 1 being at least a contributing factor, but 45s of post-fix uptime is nowhere near a rigorous disproof.

Observed, deliberately not touched (out of scope: not a crash, not wrong behavior)

• feature/damage.lpc:535 (and two identical copies, feature/damagedie.lpc:217, u/dianel/damagedie.lpc:199): DEATH_ROOM->start_death(this_object()) calls a function that doesn't exist anywhere in this archive (/d/death/gate.lpc has no start_death(), grepped the whole lib). Confirmed harmless: a call_other() to an undefined function is a silent no-op on this driver (confirmed — zero debug.log/stdout trace during an actual live death), and the real reincarnation trigger already happens via a completely different path (d/death/npc/pang.lpc's init()-triggered call_out chain, fired automatically when the ghost is moved into the room). Dead/vestigial code from an earlier version of the death room, not a functional bug — left alone per this pass's scope (programming bugs that actually crash or misbehave only, not speculative cleanup).
• One unrelated caught error surfaced ambiently during the same session, from a completely different, unvisited zone: *Illegal to move or destruct an object (/d/moon/obj/luoyi#346) defining actions from a verb function(give teapot to fei zei) in object(/d/qujing/wuzhuang/npc/mingyue#144) which returns zero. — this fired from ambient NPC/quest simulation in 五庄观 (d/qujing/wuzhuang/), a late-game zone this playthrough never visited. Not chased — out of the tested path, and root-causing it properly would need a separate deep pass through that zone's own quest scripts.
• d/ourhome/kedian.lpc:20 — the "exits" mapping's sizeof() == 2 comment doesn't match its actual single "south" key. Stale comment from a prior edit, not a functional bug (the room's actual exits work correctly as coded) — left alone.

WASM 修复摘要（迁移自 meta.json 的 group_note）

仙侣情缘 XLQY 基础版（2007）。

§7.86 跨库扫描修复（留言板 post 崩溃）

• BBS_BOARD、BULLETIN_BOARD inherit + 多余 replace_program() 致命形状（AGENTS.md §7.86，post 命令崩溃）：全档案 40 处命中，已删除多余的 replace_program(...) 调用（保留 inherit），逐文件保留原有行尾格式（CRLF/LF 按文件原样）。本次为跨库 §7.86 扫描修复（触发原因：该 bug 已在 6+ 个互不相关的血统家族独立确认，属于近乎普遍的拷贝粘贴模式），仅做编译检查（驱动干净启动、端口正常监听），未做完整 §10.7 深度游玩测试。

深度功能测试第二轮 / Deep functional test round two (2026-08-15, post driver-upgrade re-test)

驱动于 2026-08-12 升级后的重测。标准检查清单发现并修复四处问题：

1. config.fluffos：maximum evaluation cost 从 500000（已知 风险区间）提升到 5000000。 2. adm/obj/master.lpc::log_error()（AGENTS.md §7.10 大小写坑的 又一实例）：判定用的是 strsrch(message, "warning: ")（小写 w，且要求紧跟一个空格），而本驱动实际吐出的是大写 Warning: Unused local variable 一类文字，逐字节对不上，等于 完全没有过滤——每次惰性编译遇到警告都会把原始编译诊断广播给连 线玩家。改成不含开头 w/W 的 "arning:" 子串匹配（§7.10 记录的 标准写法，同时兼容大小写两种历史形态）。 3. adm/simul_efun/file.lpc：log_file() 没有 assure_file() 目录预建保护，补上调用及前向声明；cat() 补上 read_file() || "" 空值防护。 4. obj/user.lpc::reconnect()（AGENTS.md §7.108，第十五条独立确 认的血统）：adm/daemons/logind.lpc 有同款 exec(old_link, user); 踢掉重复登录写法，reconnect() 缺少 enable_commands()。按 §7.108 记录的写法预防性修复，现场用两个 真实连线复现"保持第一个连线不断开→第二个连线登录→答 y 踢掉旧连 线"验证：score 修复后立即正常显示完整角色档案。

cmds/wiz/update.lpc（§7.106）已经是正确写法，无需改动。

现场验证摘要

驱动干净启动，管理员 fluffos/Mud@2026 登录（GB/BIG5 选择→未成 年人关卡"no"→id+密码）确认 您的系统权限目前是：(admin)， update /adm/daemons/logind 成功验证真实写入权限，现场未观察到编 译警告泄漏（确认 log_error 修复生效）。踢掉重复登录重连路径现场验 证通过（见上）。debug.log 全程干净（876 行，无真实错误）。

本轮修改的文件

• config.fluffos
• work/adm/obj/master.lpc
• work/adm/simul_efun/file.lpc
• work/obj/user.lpc

§7.100 sweep (2026-08-19)

Fixed the corpus-wide inherit ROOM; ... replace_program(ROOM); redundant-replace bug (AGENTS.md §7.100). 337 live occurrences deleted: 336 via scripted sweep (fix_710_room.py), plus 1 hand-fixed roommaker-tool template (obj/roommaker.lpc, simple string-builder variant). 5 already-commented-out instances left untouched. No real .lpc source found under work/data/. Verified via build-debug driver boot: clean compile, port 40022 listening, zero new "cannot replace"/"cannot bind" debug.log lines.

§7.30 uninitialized-mapping accessor sweep (2026-08-20)

Corpus-wide mechanical sweep of the feature/skill.lpc shared-lineage bug (confirmed independently on xiakexing2017/jqxz2015/haiyang2 via round-four testing): 3 accessor(s) in this file returned a raw never-initialized mapping instance variable (defaults to int 0, not ([]), until first assigned), crashing any unguarded keys()/sizeof()/indexing caller for a fresh/untrained character. Fixed at the accessor level (mapp(x) ? x : ([])) per the documented remedy. Verified via lpcc --batch static compile check only (not a live boot) as part of a large mechanical sweep; not individually functionally re-tested live on this lib.

AGENTS.md §7.19 fix: enable_player() reentrancy from init()

feature/command.lpc's enable_player() (wrapper around enable_commands()) was reachable from an NPC's init() (via the shared std/char.lpc setup() chain), and enable_commands() is only safe to call from create() -- calling it again on an object already living() makes the driver re-invoke that same object's init() as a side effect, which recursed back into enable_player() on the same call stack until "Too deep recursion" aborted the boot on a room's first-ever visit. Fixed with a true reentrancy flag (in_enable_player_now, set at entry, cleared before every return), NOT a living()-gated guard -- disable_player() in the same file legitimately re-calls enable_commands() while already living() (sleep/wakeup via cmds/std/sleep.lpc, revive via feature/damage.lpc), which a living() guard would silently break. Verified via lpcc --batch single-file compile check (PASS). Part of the corpus-wide §7.19 sweep (Batch C).