jhfy — 江湖风云

Archive: 江湖风云.rar (#59, byte-identical duplicate: 江湖风云 (1).rar, already listed in AGENTS.md's Duplicate archives table — only this one processed; confirmed via md5sum before extracting, 5607c63bab7ebe67dd44896b145ccda7 for both). Port: 40053. Status: done (boots clean, full registration flow verified end-to-end three times across two driver sessions, with three different real Chinese names, all reaching the actual game world).

What this is

The archive extracts to raw/Jhfy/, a Windows MudOS 0.9.20 bundle (mudos.exe, mudos.dll, mud.log) with the actual mudlib root at Jhfy/world/ (confirmed via Jhfy/config.jh's mudlib directory : ./world / master file : /adm/obj/master directives — the archive's own config file, not a guess). config.jh's name field decodes (GB18030→UTF-8) to "江湖风云单机" ("Jianghu Fengyun — standalone/single-player build"), matching the archive's own filename — no slug rename needed, jhfy is a direct, accurate transliteration. ~8,979 raw files, 8,071 .lpc after the rename — a mid-sized lib for this batch.

Layout: adm/{daemons,obj,simul_efun,etc,tmp} (not adm/single/).

Lineage: a genuine third fork of the 夕阳再现 ("Sunset Reappears") family

The task brief flagged that two already-done libs in this repo — xyzxfk (#46, 夕阳再现-疯狂江湖.rar) and wmkj (#55, 未明空间.rar, live banner "江湖风云之夕阳再现") — both show a "江湖风云" name somewhere. Checked via md5sum on the three core files, per AGENTS.md's standing "verify via diff, never assume from title alone" rule, before doing anything else:

| file | jhfy | xyzxfk (#46) | wmkj (#55) | |---|---|---|---| | adm/simul_efun/chinese.c | 961d77af... | 961d77af... (same) | 961d77af... (same) | | adm/obj/master.c | 8841d611... | 861c587c... (different) | 8841d611... (same) | | adm/daemons/securityd.c | d1bd5477... | d1bd5477... (same) | a33cb521... (different) | | adm/daemons/logind.c | 24ef0ce8... | dbc4a7af... (different) | efd6e03e... (different) |

This is a genuine three-way shared lineage, not a coincidence of title: this lib's chinese.c is byte-identical to BOTH #46 and #55; its master.c is byte-identical to #55's; its securityd.c is byte-identical to #46's; only logind.c (and the room/world content) is independently maintained in each fork. This lib's own logind.lpc banner literally reads "夕阳再现" / 之 / 「江湖风云」 — i.e. this archive is very plausibly close to (or is) the actual upstream "夕阳再现之江湖风云" build that both #46 and #55 forked from/around, given it shares the *most* files (2 of 3 core files matched, vs. 1 of 3 for each of the other two). Confirmed further during the lpcc sweep: d/quanzhen*/hudi{4,5,6}.lpc, d/player/fyue_room.lpc, and several u/fyue/huashan/* files are the exact same room content (same filenames, same pre-existing typos) already documented in wmkj's NOTES.md — genuine shared world-content, not independently authored.

All of #46's and #55's proven fixes were ported/re-verified directly against this lib's actual source rather than rediscovered from scratch, per the task brief's instruction — see below for exactly which applied, which didn't, and how each was confirmed.

Fixes applied

1. AGENTS.md §15h (standard shape, ported from #46/#55 — verified this lib's chinese.c is the identical file, so the identical fix applies): adm/simul_efun/chinese.lpc's is_chinese() — GBK lead-byte range check (strlen(str)>=2 && str[0]>160 && str[0]<255) rewritten to a CJK Unicode codepoint check (strlen(str)>=1 && str[0]>=0x4e00 && str[0]<=0x9fff). 2. AGENTS.md §15h, check_legal_name() in adm/daemons/logind.lpc (this lib, like #55, has no separate named.lpc — check_legal_name is defined directly in logind.lpc, and its shape is NOT byte-identical to #55's copy since logind.c is one of the files that differs across this family — read and fixed independently, not blind-ported): - Byte-count bound strlen(name) < 2 || > 10 → character-count bound < 1 || > 5 (message text "必须是 1 到 5 个中文字" already stated the correct intended character count). - Sliding check i%2==0 && !is_chinese(name[i..<0]) (alternating GBK-lead-byte offsets, variable-length tail slice) → !is_chinese(name[i..i]) on every index (one whole character per index under UTF-8). 3. AGENTS.md §15p: adm/etc/preload had /adm/daemons/network/dns_master live — commented out proactively before the first boot attempt (CRLF-terminated file, edited carefully to preserve line endings elsewhere).

Confirmed NOT needed (verified by reading source directly, not assumed)

• §4 (master.lpc load_object()-during-compile recursion in valid_read/valid_write): read the full master.c before the first boot — both applies only ever call find_object(SECURITY_D), never load_object(). No lazy-load recursion possible.
• §7 (missing get_root_uid()/get_bb_uid()): both present and correct (return ROOT_UID; / return BACKBONE_UID;).
• §15l (master.lpc create() destructing SIMUL_EFUN_OB): this lib's master.lpc create() is a single write() line, nothing else.
• §15 / nitan-family shared-simul_efun dbase architecture bug: feature/dbase.lpc implements real local set/query/delete/add (+ _temp variants) as genuine functions (inherited via F_TREEMAP), and the simul_efun fragment adm/simul_efun/wizard.lpc defines only wizhood()/wiz_level() — no bare set/query/delete simul_efuns at all. Confirmed architecturally correct (same family as rzrmud/ wmkj), not the bug.
• §15b family (addn, remove_ansi, noansi_strlen, B2G, db_affected, clr_ansi, changed_match_path, query_bandwide, query_shadowed, tail()): grepped the whole lib for each — zero hits except remove_ansi, which IS locally defined (not a simul_efun gap) in both files that call it (u/fyue/score1.lpc, cmds/wiz/score1.lpc). message_combatd (a real gap in #55) is only referenced from commented-out code here (kungfu/skill/{yuxiao-jianfa/tianwai,tiangang-zhi/lingkong}.lpc) — confirmed dead, not a live call site, so not fixed.
• §14 (2-arg valid_override): this lib's master.lpc has the old 2-arg signature. Checked every #included simul_efun fragment (atoi, chinese, gender, object, path, wizard, message, file) for an efun::destruct/efun::move_object override written inside a fragment — the only efun::destruct override in the whole composition is in adm/obj/simul_efun.lpc itself (the real SIMUL_EFUN_OB, which already passes the existing check's file == SIMUL_EFUN_OB branch). Gap is real but provably unreachable; left as-is (matches #55's identical finding).
• §15n/§15o custom-securityd func-discrimination gap: securityd.lpc (byte-identical to #46's) IS a genuinely custom ACL (trusted_read/ exclude_read mappings keyed by directory + wizard-status), but its valid_read already has the same escape hatch documented safe in #55: if (!sscanf(base_name(user),"/clone/user/%*s") && author_file(base_name(user))==ROOT_UID) return 1; — any read attributed to a ROOT-owned system caller is allowed unconditionally. Confirmed no Read access denied anywhere across three full interactive test sessions' debug.log.
• §8d/§15o <local.h>-next-to-its-user / mid-connection #include gap: convert_lib.sh's automated pass already converts every bare (no-slash) angle-bracket local include to quoted form (converted 50 local angle-bracket includes to quotes — logged during conversion), and quotes resolve against the including file's own directory unconditionally, with no VM-context dependency — meaning this class of bug is fixed at the mechanical-conversion stage, not something master::get_include_path() needs to paper over. Verified zero remaining bare <x.h>-with-no-slash includes anywhere in work/ after conversion (grep came up empty), and zero absolute-path-in-angle-brackets (§15t item 1) or ..-relative includes (§15t item 2) anywhere in the lib either. master.lpc has no get_include_path() apply and none was added — nothing in this lib needs it.
• §8f (TYPE * name1, name2; array-modifier-doesn't-propagate bug, Dead-Souls lineage): 107 lines matched the shape via grep; spot-checked 3 (feature/equip.lpc:9's type, feature/alias.lpc:13's several, kungfu/skill/kumu-shengong.lpc:11's master) — all genuinely used as plain scalars later in the same function, correctly-written C-style declarations. This bug class does not recur in this lineage (confirmed again, matching #55's identical finding); left unfixed.
• §8e (tail() non-efun): zero call sites (grep -rn '\btail(').
• §15j (per-IP registration throttle): grepped adm/daemons/band.lpc for IsTimeAllowed/NewIps-shaped anti-flood logic — not present; no throttle to interfere with repeated registration testing.

Registration-flow verification (the critical check)

Read adm/daemons/logind.lpc's full callback chain (logon → get_id → confirm_id → get_name/get_resp → new_password → confirm_password → select_gift/set_gift/get_gift → get_email → get_gender → init_new_player/enter_world) before scripting any test. No hidden pre-id BIG5/student/client-version gate — the very first prompt ("请输入您的英文名字：") really is get_id, gated only by check_legal_id() (3-14 lowercase English letters only — no digits, confirmed the hard way, see below) and a banned_id substring check (admin, arch, wizard, apprentice, immortal, player, fuck) plus a literal "guest" ban. A valid non-empty Chinese name is accepted with no y/n confirmation step — straight into the password prompt.

Gotcha hit and documented: a first attempt used id qinfeng2 (digit in the English id) — check_legal_id() only allows a-z, so it was rejected, and every subsequent scripted --send line landed on the still-active "please re-enter your English name" prompt, cascading into a confusing multi-line rejection sequence (exactly the class of gotcha AGENTS.md warns about: re-verify with a clean id, don't assume the Chinese-name fix is broken from a cascaded transcript). Confirmed by re-running with an all-letters id.

Test 1 (first successful full run) — python3 scripts/mudclient.py 127.0.0.1 40053 --timeout 30 --idle 2 --send "qinfengsan" --send "y" --send "秦风三" --send "test1234" --send "test1234" --send "0" --send "y" --send "[email protected]" --send "m" --send "look" --send "quit":

• qinfengsan accepted as new id → confirm prompt → y → Chinese-name prompt.
• Real Chinese name "秦风三" (3 characters) accepted on the first try (no rejection loop) → password set → gift selection (0=random, y=accept) → email accepted → gender m accepted → init_new_player/ enter_world ran → landed in "客店" (Inn) room, one of the 4 start_room entries → chat broadcast confirmed correctly-rendered Chinese: 听说又来了一位叫做秦风三的少年侠士. look and quit both worked (dropped starting cloth/token items, printed "欢迎下次再来！").

Test 2 (independent, after all content fixes below were applied and the driver was restarted — LPC objects don't recompile from an on-disk edit alone) — same script, id linfeng / name 林风 (2 characters), gender f: accepted through the identical full chain, landed in a different start room this time (/d/xingxiu/beijiang, "北疆小镇" / Northern Frontier Town — confirming the random(4) start-room pick works), correct gender-appropriate broadcast (听说又来了一位叫做林风的小姑娘), look/ quit both worked.

Test 3 (final re-verification after ALL fixes, on a fresh driver restart) — id qinfengsi / name 秦风四 (3 characters), gender m: accepted immediately again, landed in "铁枪庙" (Iron Spear Temple) — a third distinct start room — quit worked cleanly.

All three sessions' debug.log grep clean for error/denied/ segmentation/Undefined function beyond the two expected benign boot-time echo lines (mudlib error handler : 1, a config-flag echo, not an actual error) and the intentional live log_error()-to-player mechanism surfacing harmless Unknown #pragma, ignored / Unused local variable compiler warnings to whoever triggers a lazy compile (original mudlib behavior, not a bug introduced here).

Driver process management note: the very first boot (launched via a plain nohup ... & disown from a Bash tool call) was killed by an unexplained external SIGTERM sometime between two tool-call turns, even though ps confirmed it alive and had already served one full successful test in between — strongly suggests the sandbox's per-tool-call process group cleanup can reap a background process that isn't in its own session. Fix: launch with setsid nohup ... & disown (own session id, ps -o pgid,sid confirms PGID==SID==PID) — the driver then survived cleanly across every subsequent tool call for the remainder of testing.

All driver processes were killed (kill -TERM) after each round of testing; no driver process is left running on port 40053.

lpcc_check.sh sweep

8,071 files. Host memory stayed healthy throughout (free -h: 3.7-4.2GB free / ~3.6GB used out of 23GB at every check — nowhere near AGENTS.md §6b's OOM-risk territory; other agents' concurrent conversions for archives #58/#61 were also running on the same host at points during this pass without issue).

| Pass | Result | |---|---| | 1 (pre-fix) | 7933 / 8071 = 98.3% | | 2 (+ 20 file fixes, see below) | 7953 / 8071 = 98.5% | | 3 (+ 8 more file fixes) | 7959 / 8071 = 98.6% (final) |

Content bugs found and fixed via the sweep (all pre-existing, confirmed

against raw pre-conversion bytes where relevant — not introduced by this pipeline)

• 6 files, "closing heredoc tag merged onto the preceding text line" (@LONG ... LONG needs LONG on its own line): d/heimuya/npc/tang.lpc, d/heimuya/tang.lpc, d/quanzhen/manglin{1,2}.lpc, d/quanzhen_old/manglin{1,2}.lpc — inserted the missing newline before the closing LONG tag. (First automated attempt via a blind regex also matched the *opening* @LONG tag and incorrectly split it into @\nLONG, breaking the heredoc start — caught and corrected before re-testing; a reminder that even a "safe-looking" regex bulk-fix across a handful of files needs a diff-check, per AGENTS.md §15t's own precedent.)
• 4 files, missing opening quote before a set("long", <bare text> block (§10-shape, exact same room content as documented in wmkj's NOTES.md — same filenames, confirming shared world content): d/quanzhen/hudi6.lpc, d/quanzhen_old/hudi{4,6}.lpc (missing opening quote only), d/quanzhen_old/hudi5.lpc (missing both opening and closing quote/\n), d/player/fyue_room.lpc (missing opening quote and closing \n").
• 1 file, invalid-GBK-lead-byte-eaten-adjacent-newline (the iconv -c gotcha from AGENTS.md's Encoding section, confirmed via raw-byte inspection: xxd on the pre-conversion file shows a genuine invalid GBK lead byte 0xa3 immediately followed by a literal 0x0a newline at the exact failure point): d/heimuya/shenggu.lpc — re-inserted the dropped newline before the closing LONG heredoc tag (the single fullwidth character that was consumed alongside the newline could not be recovered, but the sentence reads complete without it).
• 1 file, entire do_out()/look_out() function's string literals written completely unquoted (identical shape to the exact bug already documented for wmkj's d/city/sj.lpc — same filename, same content): d/city/sj.lpc — quoted every literal, initialized the previously-declared-but-never-assigned object me; to this_player() (the function is bound via add_action, so that's the correct actor).
• 4 files, #include __DIR_"feng.h" typo (missing the second underscore in __DIR__, an undefined token — same exact bug already documented for wmkj): d/wudang/taoyuan/tyroad{4,5,6,7}.lpc — fixed to #include "../feng.h" (confirmed feng.h lives one directory up from taoyuan/, matching sibling rooms that already correctly #include "feng.h" from d/wudang/ itself directly).
• 1 file, copy-paste inherit bug (same exact bug already documented for wmkj): clone/weapon/panguanbi.lpc (判官笔, "judge's writing-brush") did inherit PEN; ... init_pen(25); — PEN/init_pen don't exist anywhere in this archive. The file's own description ("这是一柄普通的精钢剑" — "an ordinary refined-steel sword") matches the SWORD template exactly. Fixed to inherit SWORD; ... init_sword(25);.
• 2 files, set_name() called with a bare string instead of an array for the second (id-list) argument: d/baituo/{obj,npc/obj}/bowl.lpc (set_name("海口大碗", "bowl") → set_name("海口大碗", ({ "bowl" }))), confirmed against feature/name.lpc's real signature (void set_name(string name, string *id)).
• 1 file, single-character typo breaking the very first line: d/quanzhen_old/obj/cloth.lpc started with a literal / cloth.txt (single slash, not //) — confirmed present in the raw pre-conversion bytes too (xxd), a genuine pre-existing typo, not a conversion artifact. Fixed to // cloth.txt.
• 4 files, shared root cause — a variable literally named new (§8g-shape: identical do_clone()-style sect-succession function copy-pasted across multiple sect NPC files, all with the same bug): object *inv, new; collides with this driver's new() keyword/efun (L_NEW token), producing a cascade of "unexpected L_NEW"/"unexpected L_ASSIGN"/"unexpected L_ARROW" errors at every later use. Found in d/quanzhou/npc/tang3.lpc, d/taishan/xingtang.lpc, d/taishan/tang2.lpc, d/taishan/tang3.lpc — renamed the variable to newob in all 4 (declaration + both use sites each). Two of the same four files (d/quanzhou/npc/tang3.lpc, d/taishan/tang3.lpc) also used a bare, never-#defined SAVE_EXTENSION instead of the real driver macro __SAVE_EXTENSION__ (used correctly elsewhere in this lib, e.g. logind.lpc) — fixed both occurrences in each file.
• 1 file, literal two-character "^M" text stuck mid-statement (not an actual carriage-return byte — confirmed via xxd on the raw pre-conversion file: genuinely the two ASCII characters caret+M, 0x5E 0x4D, immediately followed by a real \r\n — some past editor's visual "showing CR as ^M" display got typed as literal text into the file by mistake) plus two dangling, discarded-result comma-expression statements left over from what was probably an edited/shortened message_vision() call: d/mingjiao/banshanyao.lpc — removed the literal ^M text and the two inert orphaned expression statements (their results were already being discarded; removing them changes no behavior).
• 1 file, duplicate create() function (copy-paste leftover — the second, incomplete definition never calls init_whip()/set_name()/etc and is clearly superseded by the first, complete one): d/npc/m_weapon/weapon/m_bian.lpc — deleted the second, redundant create().

Remaining 112 failures — triaged by category, not fixed further (per

AGENTS.md §6b: fix genuine shared/impactful bugs, document the rest; this matches the same triage rigor applied on wmkj)

• 71 of 112 (63%) have no error message at all — a bare "Fail to load object". This is the exact lpcc --batch-from-a-near-empty-file-list artifact directly reproduced and documented in wmkj's NOTES.md (a from-scratch VM session needing some inherit-chain "warm-up" before certain compiles resolve correctly against a short file list, not a real bug) — not re-derived from scratch here, just cross-checked that the files in this category are the expected shape (personal wizard-zone content under u/fyue/huashan/*, various d/* room/NPC files not on the registration/boot path) and confirmed none of their names appear anywhere in any of the three clean interactive-test debug.logs.
• 8 files (adm/daemons/network/{ms,userid,pingtcp,mail_serv,netmail, telnetd,pingd,inetd}.lpc, Cannot #include {config,priv,uid,mailer, daemons,post}.h + assorted undefined variables/functions): the dead intermud/mail/telnet-daemon sibling suite that dns_master itself belongs to — confirmed none of these are referenced anywhere in adm/etc/preload (only dns_master was, and it's now commented out) or reachable from the registration path. Matches AGENTS.md §15p's "this whole daemon family is dead weight in this environment" finding exactly.
• 3 files (meskill/skill_model_{unarmed,default,weapon}.lpc, Undefined variable SKILL_CH_NAME/SKILL_BASE_NAME/SKILL_EN_NAME/action, Undefined function NewRandom): confirmed (same conclusion as wmkj) these are code templates read as plain text and macro-substituted by an in-game skill-authoring tool — the macros are deliberately never #defined anywhere, by design, meant to be filled in by the generator at runtime, not compiled directly.
• 1 file (adm/obj/simul_efun_bak.lpc, Invalid simulated efunction override): a _bak suffix backup copy of simul_efun.lpc (alongside adm/obj/master.c.bak/.org, also unused) — not referenced by config.fluffos's simulated efun file directive, dead by construction.
• 2 files (feature/vendor_sale.lpc, its inheritor kungfu/class/npc/xiaoer.lpc): a large, genuinely pre-existing syntax breakage — confirmed byte-identical to the raw pre-conversion file (no diff), so this was already broken in the original archive, not something introduced here. The do_deposit()/do_setprice()/pay_back() functions are missing nearly every semicolon and several closing parentheses throughout ~60 lines — far beyond a single-typo fix, and this file's own NPC (kungfu/class/npc/xiaoer, a kung-fu-hall shopkeeper) is not the same file as the actual start-room innkeeper NPC (/d/city/npc/xiaoer, a completely separate, working file) that appeared live in every interactive test. Documented as a known, non-critical, pre-existing content gap — not attempted (would require guessing the author's intended statement boundaries throughout, i.e. fabricating content, not fixing a diagnosable bug).
• 2 files (d/player/ltsh/npc/saveme.lpc, its inheritor d/player/ltsh/tst.lpc, Bad type for argument 1 of exert_function (string vs int)): the exact same file, same bug, already documented and left unfixed in wmkj's NOTES.md — exert_function expects a string skill-name (confirmed against ~100+ other call sites in this lib that all pass one), but this file passes a bare int 10 with no surrounding context indicating what string was intended. Confirmed genuinely shared content (not independently duplicated) and left unfixed for the same reason as before: fabricating a guessed skill name would be inventing content, not fixing a diagnosable bug. Decorative/minor NPC file, not on the registration path.
• 2 files (d/kaifeng/east.lpc missing guanli.h, d/city/npc/ zhuziliu.lpc missing /u/fyue/job/askjob.h): genuinely missing headers — confirmed absent anywhere in the archive (§13-style content gap, not a reference-fixing bug). guanli.h missing matches the exact gap already documented in wmkj's NOTES.md for the same file family.
• 3 files (d/gumu/duanchang.lpc + the shared d/gumu/tiao.h it #includes, and its siblings d/shaolin/houdian.lpc/d/shaolin/npc/obj/ houdian.lpc which also #include it — the latter two happen to compile clean standalone despite sharing the include): tiao.h is, on inspection, not really a header at all — it's a complete standalone room definition (its own inherit ROOM; void create() {...} void init() {...}) for a "跳崖" (cliff-jump) minigame room, mistakenly given a .h extension and #included into duanchang.lpc rather than existing as its own loadable .lpc object. Since duanchang.lpc already has its own inherit ROOM/create()/init(), the #include causes "Illegal to redefine 'nomask' function _set/_delete/_query" (via the doubled ROOM→F_DBASE inherit chain) and "Redeclaration of function create/init". Left unfixed: untangling which parts of tiao.h were meant to be inherited *behavior* (the do_tiao/do_yes/ do_no/win/fail cliff-jump functions) versus which were leftover *room* boilerplate that should never have been in a #included fragment requires a design decision this pass isn't positioned to make safely without either dropping real content or guessing wrong; decorative side-content, not on the registration/boot path.
• A handful of one-off pre-existing gaps, each confirmed non-critical and not chased further given diminishing returns (all are admin tools, personal wizard-zone content, decorative kung-fu skill files, or doc reference material — none appear in any of the three clean interactive test transcripts): doc/mudlib/efuns.lpc (a plain-text efun-name reference list, not code, caught by the blanket .c→.lpc rename — same AGENTS.md §12 shape as the shenminmap.c precedent) and doc/mudlib/efuns/shiwu.lpc; quest/murong/qlist3.lpc; clone/misc/ objmaker.lpc (a builder tool); clone/topten/gaoshou.lpc (missing GAOSHOU macro); d/city2/npc/zhengmen.lpc (unbalanced braces); d/npc/master.lpc (a bare set("inquiry", ...) fragment with no create()/inherit at all — looks like an NPC-dialogue-table fragment meant to be #included elsewhere, but no other file references it by this exact path; likely orphaned); d/huashan/npc/doctor.lpc + its inheritor d/huashan/doctorroom.lpc; d/huashan/map.lpc and u/fyue/ huashan/map.lpc (both pure ASCII-art room-layout maps, not a line of LPC — same §12 shape as doc/mudlib/efuns.lpc, grep -rl for their basenames elsewhere comes up empty); d/shaolin/npc/obj/shuitan.lpc (one undefined variable, ging_cost, likely a typo'd variable name); d/quanzhen_old/npc/haodatong.lpc; u/snow/who.lpc and u/fyue/user.lpc (personal wizard-sandbox admin tooling); cmds/usr/whotest.lpc (a "test" command, name suggests experimental/dead); cmds/bos/socket.lpc (dump_socket_status, a debug efun with no equivalent in this driver's socket package); and several kungfu/skill/*.lpc one-off syntax errors (iceforce/chillgaze, hanbing-mianzhang2, huang, xixi-unarmed, quanzhen-xinfa) — decorative combat-skill content, none reachable from the registration flow or the interactive test's actual path.

Encoding

convert_lib.sh's automated pass: 8,649 converted cleanly, 295 already-UTF-8, 33 lossy (invalid bytes dropped via iconv -c fallback — mostly NPC/room/data files, e.g. d/quanzhen_old/obj/bookshelf.c, u/fyue/huashan/{map.c,intro}, cmds/usr/{nick.c,whotest.c}), 2 skipped as genuinely binary. One of the lossy files (d/heimuya/shenggu.c) was confirmed via xxd to have a genuine invalid-GBK-lead-byte-then-newline sequence at the exact failure point — the specific "iconv -c eats an adjacent real newline" gotcha from AGENTS.md's Encoding section, fixed as described above. No un-converted GBK stragglers found (re-ran the AGENTS.md straggler check: file -b reports text/script/empty for every .lpc/.h in work/).

Config

config.fluffos adapted from Jhfy/config.jh (labeled "MudOS 0.9.20", GBK→UTF-8 converted before any other edit, per AGENTS.md §5). Original port number : 6666 replaced with 40053 (this project's port assignment for archive #59, per the task brief — next free port after 40051-40052 reserved for archives #57-58). mudlib directory points at /home/sunyc/src/mudlib/libs/jhfy/work (absolute path). All other directives carried over, matching the tuning-knob template used for the last several libs in this batch (wmkj/xiakexing3 shape). log/ directory created as a sibling of config.fluffos (not inside work/), driver always launched via cd libs/jhfy && setsid nohup .../driver config.fluffos (see the "Driver process management note" above for why setsid specifically was needed in this session).

2026-07-23 re-verification pass (driver rebuild + formatter + WASM)

• LPC reformat: ran format-corpus.mjs over all of work/ (8,071 .lpc files) — 8,004 written, 36 already-idempotent, 31 refused (self-checked round-trip failures on messy legacy syntax, expected).
• Native retest against the freshly-rebuilt driver: booted clean on port 40053, zero fatal preload errors. Full registration flow with a fresh real Chinese name ("秦风九"/qinfengjiu) through the complete wizard (id → confirm → Chinese name → password → confirm → gift roll → email → gender) into the actual game world (客店); look (room description correct, shown consistently), score (full character sheet correct: attributes/天赋/HP-food-water bars/potential all populated) and quit (correct item-drop-on-quit flavor text + clean disconnect) all verified with real output. No runtime errors in the session, no regressions from the reformat or the new driver binary.
• WASM test (scripts/wasm_client.js against build-wasm/src): boots cleanly, same preload warnings as native, no fatal preload errors. Login is blocked immediately on connect by the documented query_ip_number() WASM limitation (see AGENTS.md): adm/daemons/ logind.lpc's logon() does str = query_ip_number(ob); if (BAN_D->is_banned(str) == 1) { ...destruct(ob)... } — under WASM query_ip_number() doesn't return a well-formed dotted-quad, so BAN_D's ban-list matching misfires and every connection is rejected with "你的地址在本 MUD 不受欢迎" before ever reaching the id prompt. Confirmed via direct source read as the documented driver-side gap, not a mudlib bug — not patched. Marking this lib as "boots under wasm, login gated by the documented IP-check limitation" per the task's guidance, not "fails under wasm."

WASM-enablement pass (2026-07: loopback-allow + throttle exemptions + admin seeding)

Standard pass per AGENTS.md §1.3b/e + §1.5. Gates found and patched:

• adm/daemons/band.lpc:106 is_banned(): short-circuit return 0 (allow) for loopback/empty/malformed site strings BEFORE the original sscanf(site, "%s.%s.%s.%s") != 4 => return 1 (banned) check. This is the exact gate that made this lib "WASM 下无法登录" (README): under WASM query_ip_number() returns a malformed string, the dotted-quad sscanf fails, and every connection was rejected as banned. With the short-circuit, malformed == local == allowed; the anti-virtual-IP (.0/.255) and regexp ban-list logic below is unchanged for real remote dotted-quads.
• adm/daemons/logind.lpc:72 new helper is_loopback_conn() (declared :35); used to exempt loopback from:

- :173 — ban_cnt > 16 cap on concurrent same-IP pre-login connections in logon(); - :235 — ip_cnt > 8 same-IP player cap in get_id() (the non-welcome/non-netclub branch).

• No uptime() startup-grace gate exists in this lib (all uptime() uses are content randomness).

Admin account: fluffos / Mud@2026 / 浮浮, registered through the real native flow (id → y → Chinese name → password ×2 → gift 0/y → email → gender m). Granted (boss) — NOT (admin) — via adm/etc/wizlist (added fluffos (boss) alongside the original fyue (boss)): this lineage extends the ES II wiz_levels ladder with (president)/(admin)/(boss)/(ceo) above (arch), and (boss) is the top rank the securityd ACL tables actually trust (trusted_read/ trusted_write "/"), as well as what the original admin fyue holds. Verified after restart: login shows 您目前权限：(boss), update /cmds/usr/score recompiles successfully. Save files (must be committed): work/data/user/f/fluffos.o, work/data/login/f/fluffos.o.

Retest: fresh registration (秦风/ceshizhe) end-to-end + look/score/quit clean; fluffos login + update clean; debug.log free of errors; test character saves removed.

Dual-mode verification pass (2026-07-24)

• Native (port 40053): fresh registration ceshier/秦风 end-to-end (id → y → name → password ×2 → gift 0/y → email → gender m) into 客店; look/score/quit all correct. fluffos/Mud@2026 login shows 您目前权限：(boss) and update /cmds/usr/score prints 成功. debug.log: zero errors. Driver killed by exact PID.
• WASM (build-wasm with query_ip_number/resolve fixes): the previous blocker — BAN_D rejecting every connection on the malformed-IP sscanf — is GONE, exactly as predicted: the enablement pass's band.lpc:106 loopback/malformed short-circuit plus the driver-side query_ip_number() fix both cover it. Fresh registration wasmceshi/秦风 end-to-end into the game world (北疆小镇 starting branch), look/score/quit all correct; second invocation: fluffos admin login (boss) + update 成功. Verdict: native OK + wasm OK. No new fixes needed this pass. Test character saves removed; fluffos.o timestamp churn reverted.

WASM 修复摘要（迁移自 meta.json 的 group_note）

此前被错误标记为某个不存在于本项目任何档案的原始压缩包文件名的 duplicate_of；这里的内容其实是完全独立、可游玩的游戏（有自己的 -N 变体编号、自己的端口、自己的 README）——duplicate_of 已清除。

深度功能测试（§10.7，本轮）：直接核对手足档案 jhfy2 已确认的 bug

jhfy2（本项目里的另一个"江湖风云"变体）此前的 §10.7 深度测试发现 了 3 个 bug（printf 调试残留、§8.9 食物/饮水年龄检查错对象、2 处 §7.68 死亡软锁）。这次直接核对了 jhfy 是否也带着这几个 bug，而不 是从零排查：

• §8.9 命中：adm/daemons/logind.lpc 的 enter_world() 里 if (ob->query("age") == 14) 同样检查了错误的对象（应为 user->query("age")），已改正。
• §7.68 命中，2 处：d/death/npc/{bgargoyle,wgargoyle}.lpc 的 death_stage() 都有把"角色永久离开"和"暂时不在场"混在一起的 if (!ob || !present(ob)) return; 写法，按各文件自己的 5 秒重试 间隔拆分成"暂时不在场则重试"。也确认了 REVIVE_ROOM （/d/city/wumiao）和 DEATH_ROOM（/d/death/gate）两个宏指向 的档案都真实存在。
• printf 调试残留：未命中——adm/daemons/logind.lpc 里所有 printf(...) 调用都是正常的玩家可见格式化输出（用法类似 write(sprintf(...))），没有 jhfy2 那种裸 printf("%O\n", ob) 物件引用泄漏，确认后未做改动。

用真实驱动完整验证了一次全新注册（英文 id→中文名字确认→密码→天赋 点数（随机 0）→email→性别，比 jhfy2 多了"是否满意随机中文名字"这一 道确认），进入"客店"后 score 正确显示食物/饮水两条状态全满、属性 数值和分配一致；移动到"南大街"实际攻击 NPC"小贩"，角色被打死，干 净地触发"你死了"→"鬼门关"死亡场景→NPC"白无常"自动完成整套复活对 白（不需要玩家任何操作）→魂魄归位→正确送达 REVIVE_ROOM（"武庙"， 和宏定义一致）——debug.log 全程保持空白，确认三处修复全部生效， 没有任何遗留问题。

未继续测试的部分

时间关系，没有测试拜师、购物、留言板。

更正（2026-08-05）：§7.68 复活软锁"修复"已撤销

上面提到的"鬼魂离开/不在场时被永久放弃复活流程"曾被当作 AGENTS.md §7.68 记录的一类 bug 修复（把单次判定改成每 5 秒重试）。经用户指出并 重新审视：这更可能是有意的游戏设计，不是 bug——大多数这类档案里 鬼魂根本无法自行移动，所以"不在场"要么从未真正发生，要么是"离开去 在阴间游荡，想回来时再走回这个房间、流程会通过 init() 重新从头开始" 这种有意为之的宽松机制，而不是需要强制追上玩家的错误。强行重试还可能 引入新问题：如果鬼魂之后又走回这个房间，旧的重试和 init() 重新触发的 新一轮流程可能同时运行，导致对话重叠错乱。已把这处改动撤销，恢复成 原始的 if (!ob || !present(ob)) return; 单次判定写法（bmxkx2001 除外——那份档案里这确实是一个真实存在、经过实际复现验证的 bug：鬼魂 本身完全无法移动，是另一个不相关的 NPC 强行把鬼魂拖走导致的）。详见 AGENTS.md §7.68 顶部的撤销说明。

§7.86 跨库扫描修复（留言板 post 崩溃）

• BULLETIN_BOARD inherit + 多余 replace_program() 致命形状（AGENTS.md §7.86，post 命令崩溃）：全档案 93 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（2026-08-13，round two，新驱动重测）

Re-tested against the freshly-rebuilt build-debug/src/driver（post 全库 quest_times/win_times %-operator 修复 + Warning/warning 驱动文本回退）。log_error()（adm/obj/master.lpc）已经在更早一轮 （引用注释里的"AGENTS.md #15w"）正确修复过，本轮只发现并修复了 log_file() 一处。

发现并修复的 PROGRAMMING bug

1. log_file()（adm/simul_efun/file.lpc）完全没有 assure_file() 保护（AGENTS.md §7.11-class 的又一确认实例，与本轮 ffxymud/ hc 完全同一形状）：adm/daemons/logind.lpc 的 get_gender()（新角色注册流程的最后一步）紧跟着调用 log_file("login/newid.log", ...)——LOG_DIR 下的 login/ 子 目录若不存在，会在每一个全新角色注册完成的那一刻未捕获抛出，属 于会影响核心注册路径的高优先级实例（不只是 nosave/* 这类只在 管理指令才会触发的低优先级路径）。已补上 assure_file(LOG_DIR + file);（含前向声明）。

Proactive checks（无需改动）

• win_times 修复确认存在且正确：d/city2/npc/refereew.lpc:146。
• 未发现 message() simul_efun 包装函数——不适用 message()-missing-varargs 这一类 bug。
• 管理员账号（fluffos/Mud@2026，adm/etc/wizlist 里已有 fluffos (boss)）此前已用真实注册流程创建并提交，本轮复用，只做 真实登录 + update 复测：score 确认"您目前权限：(boss)"， update /adm/simul_efun/file（就是本轮改过的文件）确认可正常重 新编译。log/debug.log 时间戳全程未变化（Aug 5，早于本次会 话），确认无新增未捕获运行期错误。登录本身产生的存档时间戳类微 小 diff（data/{login,user}/f/fluffos.o 的 last_on 字段）已用 git checkout 撤销，不提交。驱动最终按精确 PID kill，ps -p 确 认已退出。

（data/{login,user}/c/chenba.o 是此前会话遗留的未提交测试存档 ——Aug 5 mtime，早于本次会话，未受本轮任何操作影响，未触碰。）

§7.100 房间基类 replace_program() 扫尾修复（2026-08-19）

ROOM 宏（/inherit/room/room，见 include/globals.h）在本档案 2,245 处房间文件的 create() 里紧跟 inherit ROOM; 之后又多余调 用了一次 replace_program(ROOM);——AGENTS.md §7.100 记录的同一个 休眠 bug（多余调用给对象打上永久"pending replace"标记，对象一旦绑 定任何闭包就会崩溃）；本档案属于 XYZX/fyue 房间生成工具血统， clone/misc/roommaker.lpc/u/fyue/misc/roommaker.lpc 两份逐字节 相同的副本。用 fix_710_room.py 扫过 work/，删除 2,243 处标准 形状；两份房间生成工具各剩 1 处字符串拼接变体 str += "...replace_program(ROOM);..."（heredoc 模板里的独立行 已被脚本正常扫到），手工改成 str += "\n\tsetup();\n}\n";。修复 后 work/ 下 0 处存活 replace_program(ROOM) 残留，105 处转档之 前已注释掉的 // 行原样保留，work/data/ 下没有真实 .lpc 源码 命中。git diff --stat 显示 2243 个文件净删 2245 行，与脚本自报 数字 + 2 处手工编辑吻合。

驱动干净启动（零新增编译错误、端口正常监听、debug.log 无任何 "cannot replace"/"cannot bind"行），巫师账号 fluffos/Mud@2026 确认"您目前权限：(boss)"后 look/goto 走读了 2 个曾经命中过这 个 bug 的房间（u/snow/wudujiao/yaoshi.lpc、 u/snow/wudujiao/zhushe.lpc）均正常，quit 干净退出。登录存档 的时间戳增量已用 git checkout HEAD -- 撤销，未落入提交。

深度功能测试（round four, 2026-08-20）：拜师 / 购物 / 留言板 + 清单复核

环境说明：本轮会话开始时宿主机之前已重启过（archives-side build-debug/ 符号链接不存在），实际驱动二进制路径是 /home/sunyc/src/fluffos/build-debug/src/driver （不是任务提示里写的 mudlib/build-debug/src/driver——那个路径这台机器上 从未存在过，lpcc_check.sh 自己引用的也是 ~/src/fluffos/build-debug/...）， 从这个路径正常编译启动，端口 40053 正常监听。

1. 拜师（sect apprenticeship）：干净通过

attempt_apprentice() 的实际调用点在 cmds/skill/apprentice.lpc（apprentice/bai 指令），选了武当派开山祖师张三丰（kungfu/class/wudang/zhang.lpc），通过 d/wudang/xiaoyuan.lpc（"金殿"）房间的 set("objects", ...) 克隆放置定位到该 房间。管理员 fluffos 用 goto /d/wudang/xiaoyuan 抵达后 bai zhang：

☆你想要拜张三丰为师。
张三丰说道：我武当派乃内家武功，最重视内功心法。

命中 zhang.lpc 里 taiji-shengong 技能不足 50 的合理拒绝分支（fluffos 账号没有武当内功技能），机制走通、拒绝理由合理，无崩溃、debug.log 无新增 错误——按验收标准算干净通过（不需要真正收徒成功才算过）。

2. 购物（shopping）：完整验证买入 + 当铺卖出双向交易，两级货币确认正常

• 买入：确认 feature/vendor.lpc 是一份未被任何文件使用的旧副本（grep -rl 'inherit.*vendor\.lpc' 全库零命中），真正被 NPC 使用的是同目录下 F_VENDOR 宏指向的 feature/dealer.lpc（buy/sell/list/value/ pawn/redeem/check 七个函数的真正实现）。d/city2/xidan.lpc （"西单"）房间的小贩 d/city2/npc/xiaofan.lpc 只 add_action 了 buy/list（符合项目已知的"普通商铺只卖不收"内容设计，不是 bug）。 用 clone /clone/money/silver 给 fluffos 变出一两银子后：

`` ☆你向小贩买下一串冰糖葫芦。 ☆ 你身上带着的物品有(负重 4%)： 五十个铜板(Coin) √布衣(Cloth) 糖葫芦(Bingtang hulu) 法传送帖(Trans site) `` 一两银子买五十文的东西，正确找零五十个铜板——两级/三级货币（金/银/铜）换算 正确。

• 卖出（两级货币 sell-back）：inherit/room/hockshop.lpc 全库 grep -rl 零继承者，确认是真正的死代码基类（不是"可达但没测"），未强行测试。改用 实际有 add_action("do_sell", "sell") 的当铺 NPC d/city/npc/tang.lpc （通过 d/city/dangpu.lpc"当铺"房间的 objects 克隆定位），把刚买的 糖葫芦拿去卖：

`` ☆你把身上的一串冰糖葫芦卖掉四十文钱。 ☆ 你身上带着的物品有(负重 5%)： 九十个铜板(Coin) ` 五十文的东西按 do_sell()` 的 80% 折算卖得四十文（50+40=90，与背包铜板 数吻合）——买入/卖出双向真实交易全部走通，无崩溃。

3. 留言板（board posting）：确认 §7.114 不适用，真实多行留言完整验证

先静态核对 feature/edit.lpc（本档案 F_EDIT 的实际实现）：

void input_line(string line, string text, function callback) {

没有 private 修饰符——grep -n "private.*input_line" libs/jhfy/work/feature/edit.lpc 零命中，§7.114（input_to() 递归回调被误标 private 导致多行输入第二行起 静默丢失）在本档案不成立。

再用真实驱动做端到端多行留言测试：goto /d/city/kedian（"有间客栈"）， post 测试留言标题，输入两行正文，. 结束，read new：

留言完毕。
☆
[1] 测试留言标题                             浮浮(fluffos)(Thu Aug 20)
----------------------------------------------------------------------
这是第一行测试内容。
这是第二行，确认多行留言能完整保存。

两行正文全部保存并读出（不是只有第一行）——§7.86 之前修复的 replace_program() 崩溃 bug 也没有复发（post 指令本身没有崩溃）。测试帖 用 discard 1 清除，board 存档目录（work/data/board/kedian_b.o，测试前 不存在）测试后一并删除，不留churn。

4. 清单复核（静态检查，均确认无需改动）

• §7.90：config.fluffos:40 maximum evaluation cost : 5000000——已是 大数值，非原始小默认值，确认无需改动。
• §7.100：grep -rnE "replace_program\(ROOM\)" work --include="*.lpc" | grep -vE '^\S+:[0-9]+:\s*//' 零命中——全库残留的 100 处 replace_program(ROOM) 全部是行首 // 注释掉的死代码，确认此前的扫尾 修复完全生效，无遗漏。
• §7.111：adm/obj/master.lpc:211 standard_trace() 里 objectp(error["object"]) ? file_name(error["object"]) : "<none>"——已 有正确的 0/null 守卫，确认无需改动。
• §7.112：全库 grep -rl 'call_out("death_stage"' 只有 3 个文件 （d/death/npc/{bgargoyle,wgargoyle}.lpc、d/shaolin/npc/yu-zu2.lpc）， 三个文件的 init() 都已经带有 death_stage_active 的 query_temp/set_temp/delete_temp 重入保护——不在 AGENTS.md §7.112 记录的"残留缺口"名单（hhsj/nt6nitan6win/jh2006）里，本档案已经 是修复过的状态，确认无需改动。
• §7.113：clone/user/user.lpc 的 reconnect() 函数第 198 行确认 调用了 set_heart_beat(1)，adm/daemons/logind.lpc 的 reconnect() 正确转调 user->reconnect()——确认无需改动。
• §7.114：见上方第 3 项，已现场验证不适用。
• §7.115：include/globals.h 里唯一含 QUEST 的宏是 #define QUEST_D(x) ("/quest/qlist" + x)（一个函数式宏，不是指向单个 缺失文件的静态路径）和 #define QUEST "/inherit/quest.lpc"，后者对应 的 inherit/quest.lpc 文件确认真实存在——本档案不受影响，符合 AGENTS.md"仅 aoxiangtianji 命中"的记录。
• §7.79：grep -rn '\baddn(\|\baddn_temp(' work --include="*.lpc" | grep -v /data/ 零命中，与 NOTES.md 之前"零调用点"的记录一致，且本档 案不在 AGENTS.md §7.79 记录的 6 个真正受影响的库（xfbhh/hhsj/ nitan170911/nitan6/nt6/nt6nitan6win）名单内——确认不适用。

本轮结论

四项深度测试（拜师、购物买入、购物卖出、留言板多行发帖）全部干净通过， 七项清单复核全部确认无需改动，本轮未发现任何新的 programming bug， 未做任何代码改动。debug.log 全程只新增两行良性 Unknown #pragma, ignored 编译期警告（inherit/char/master.lpc、feature/dealer.lpc 首次被本次会话触发懒编译时输出），没有任何 error/denied/segmentation 等级别的新增内容。驱动测试结束后按精确 PID 正常终止（kill -TERM， ss -ltnp 确认端口 40053 不再监听）。测试产生的存档 churn （fluffos.o 的 last_on 时间戳、临时 data/board/kedian_b.o）已全部 撤销/删除，git status 干净（仅剩此前会话遗留、与本轮无关的 data/{login,user}/c/chenba.o，原样未动）。

§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): 4 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.