grav1tyzero/rifts-ds → riftsds

Source: git clone https://github.com/grav1tyzero/rifts-ds (cloned 2026-08-30). Mudlib root is the repo's lib/ directory; bin/, extra/ (a wolfpaw/ subdir containing classic MudOS driver-build headers macros.h/configure.h/system_libs -- driver source, not mudlib code), fluffos-2.23-ds03/ (a full bundled driver source tree), win32/, and the top-level license/readme/batch files were not converted. Slug riftsds, number 955, port 40257.

Dead Souls 3.9 lineage -- same base engine family as the already- onboarded ds386/dsI/dsII/dsIII/dshakkard/deadsouls_fluffos (§2.1 lineage confirmation: secure/daemon/master.lpc is near-byte-identical to ds386's own pre-fix raw baseline, differing only in the same static→nosave/.c→.lpc conversion artifacts). The ASCII boot banner reports "Mudlib: Dead Souls 3.9" (a later point release than ds386's 3.8.6), which explains the bundled domains/std Lima-style builder-school example area that ds386 doesn't have. Not a duplicate of the already-onboarded rifts2 (tsathoqqua/RiftsMUD2) -- that lib is DarkeLIB/Nightmare lineage, a completely different codebase that happens to use the same tabletop setting.

This repo is a live/played development snapshot, not a pristine archive. It arrived with three real pre-existing wizard accounts (gravity, thurtea, shrub -- all full admins per secure/cfg/groups.cfg), a huge number of realms/thurtea/tmp/*.lpc scratch files (leftover ed/eval-session temp files from live development, hundreds of randomly-named files), and only 10 total git commits upstream ("money pops up a lot more than I thought", "multiple fixes credits added mana now ppe", etc.) -- this reads as one developer's personal in-progress mud, not a finished release.

1. Conversion

convert_lib.sh ran normally (GB18030 default encoding -- this lib is pure English/ASCII, so the encoding pass was a no-op except for a handful of genuinely non-Chinese legacy-8-bit files, see below): 5605 already-UTF-8, 7 converted, 9 lossy, 33 skipped-binary. 3219 .lpc files after rename, 747 literal .c" references fixed, 32 local angle-bracket includes converted, 1605 files touched by the static→nosave sweep.

Lossy-conversion files, all pre-existing (not conversion fallout), all fixed:

• doc/CREDITS, doc/help/players/credits, domains/town/txt/mystery.txt -- Latin-1/MacRoman punctuation, byte- identical to ds386's own known-bad set for these exact filenames (this lineage's stock Dead Souls doc tree). Fixed by directly copying ds386's already-corrected versions of these three files (confirmed byte-identical match before copying).
• secure/sefun/translate.lpc (the "speak a foreign language badly" word-scramble table) -- one Finnish entry ("kyrp\xe4","r\xe4k\xe4") is genuine Latin-1, and GB18030 decoding silently mangled it into "kyrp,"r鋕," (the classic iconv -c-eats-an-adjacent-real-byte shape, AGENTS.md §4.1 -- here it also ate the string's own closing quote and the following comma). A second entry ("\xd0\xa5\xd0\xa3\xd0\x98\xd0\xa2\xd0\x90", Cyrillic "ХУИТА") was ALREADY valid UTF-8 but got re-mangled into other valid-but-wrong CJK codepoints by the same GB18030 pass. Both fixed by hand, matching ds386's own already-published fix for this exact byte-for-byte-shared file.
• save/intermud.o, secure/save/functions.o, realms/thurtea/tmp/ *.c (4 files): stale runtime caches (an I3-router mudlist cache, an empty function-signature cache, and 4 of the hundreds of ed scratch temp files) -- harmless, left as-is (all valid UTF-8 in their converted form, no fix needed).

2. Ported ~160 already-proven ds386 driver-compat fixes wholesale

Confirmed via a fresh from-scratch reconversion of ds386/raw (into a scratch dir, matching this session's own convert_lib.sh output byte- for-byte) that large parts of this lib's shared "engine" tree (outside domains/) are byte-identical to ds386's own PRE-FIX baseline. A Python script walked every file under secure/, lib/, cmds/, daemon/, std/, obj/, verbs/, powers/, cfg/, include/, ftp/, open/, shadows/, doc/, and domains/; wherever this lib's file was byte-identical to ds386's freshly-reconverted (unfixed) version of the same path AND ds386's own committed (fixed) version of that path differs from the unfixed baseline, this lib's file was directly overwritten with ds386's fixed version. 160 files ported this way, including the highest-impact fixes from ds386's own onboarding: secure/daemon/master.lpc (uid stubs, runtime_config.h already swapped for the driver's canonical copy), secure/sefun/ economy.lpc (the to_int()-missing float-corruption bug, AGENTS.md §7.121), lib/editor.lpc/lib/interactive.lpc/lib/nmsh.lpc (the OLD_ED single-entry-ed() rewrite, AGENTS.md §6.2 -- see below for 2 files that needed the SAME fix applied by hand since their content had locally diverged from ds386's baseline), secure/lib/connect.real (the static/array-idiom fixes for Dead Souls' self-rewriting login object), secure/include/lib.h (LIB_REMOTE stale path), lib/events/ look.lpc+look_in.lpc (the SetLong/SetInternalDesc narrow-string- type bug, AGENTS.md §7.81-class), daemon/command.lpc (the .c→.lpc filename-slice bug that broke the stock score command), and secure/sefun/sockets.lpc/instances.lpc (WASM sockets-package- absent gutting -- not yet verified under WASM here, see below).

~724 other shared-tree files diverged from ds386's baseline (this lib is a materially newer/extended Dead Souls snapshot -- new systems not present in ds386 at all: a magic-point/P.P.E. resource, parties, voting, teacher/apprentice skill training) and were left untouched by the port; each one that actually failed to compile was triaged individually (§4 below).

3. Manual fixes beyond the ported set

• The TYPE array NAME idiom (AGENTS.md §4.3/§6.3), same driver build (ARRAY_RESERVED_WORD undefined, confirmed against this driver's own options_internal.h): a quote/comment-aware Python sweep across the whole work/ tree found and fixed 230 single- declarator instances across 62 files (object array x → object *x, etc.) plus 11 multi-declarator lines (string array a, b, c; → all three get *, verified per-variable against actual usage -- explode()/foreach/index-access confirms array use -- not blindly applied): daemon/classes.lpc, daemon/help.lpc, daemon/ seasons.lpc, daemon/races.lpc, domains/town/npc/brandy.lpc, secure/daemon/players.lpc (×3 lines), secure/cmds/admins/ mudconfig.lpc, lib/magic.lpc (send_to), and lib/chapel.lpc (Religion/Deities/DeityIds, initially missed in the first pass since it wasn't in the multi-declarator report handled at the same time -- caught by the lpcc_check.sh sweep's "Bad assignment (string vs mixed*)" errors and fixed separately). New AGENTS.md §7.163: found one instance with a class compound type instead of a bare base type (lib/body.lpc's class MagicProtection array GetMagicProtection()), invisible to a plain-base-type regex.
• The .c→.lpc filename-slice-length bug (AGENTS.md §7.118-class) -- several sites where the code checks a suffix against the 4-char .lpc string but slices/truncates by the old 2-char .c length: daemon/help.lpc (3 sites in LoadIndices()'s helper closures + 1 in GetTopic()'s guard, this file wasn't part of the ported set since it had genuinely diverged from ds386's version), verbs/ builders/initfix.lpc, lib/std/boobytrap_object.lpc, secure/obj/ weirder.lpc (3 sites), secure/cmds/admins/mudconfig.lpc, secure/daemon/rooms.lpc, secure/sefun/sefun.lpc. All fixed last(x,2)/[<2..]/[0..<3] → last(x,4)/[<4..]/[0..<5] paired with the matching truncate(x,2)→truncate(x,4). Verified via a corpus-wide re-grep that zero unrelated 2-char-suffix uses (.o, "/g", ", ", "--") were touched.
• lib/std/room.lpc's SetLong wrapper was narrower than the SetInternalDesc it forwards to (string SetLong(string str) vs the already-fixed mixed SetInternalDesc(mixed str) in lib/events/ look_in.lpc) -- widened to mixed SetLong(mixed str). This broke every room using SetLong( (: SomeClosureFn :) ) for a dynamic description (confirmed live: secure/room/router.lpc throws Bad type for argument 1 of SetLong ( string vs function ) without this fix). Two OTHER SetLong overrides (lib/die.lpc, lib/ interactive.lpc) are also narrowly string-typed but are only ever called with literal strings in this codebase -- left alone per scope discipline (no error, no fix).
• lib/player.lpc's Setup() declared string oldparties = PARTY_D->GetOldParties(); but GetOldParties() is declared (and actually returns) string * -- a plain wrong-type local variable, not the array-idiom bug. This crashed RemoveExtraChannels() (Bad type for argument 1 ( string * vs string )) on literally every successful login (the function runs unconditionally in Setup()). Fixed the declaration to string *oldparties.
• domains/town/npc/zoe.lpc: SetCurrencies( ({ "universal credits"" }) ); -- a stray doubled closing quote, present in the raw archive (confirmed via raw/, not conversion fallout). Broke this NPC's compile with a cascading End of file in string, which took domains/town/room/bank.lpc down with it (Zoe stands in the bank). Fixed by removing the extra quote. Verified live: the bank room now loads correctly with "Zoe the bank teller" present.
• domains/Praxis/supply2.lpc: set_pre_exit_functions( ({ "north" }), ({ "go_north" }) ); -- the exact same broken macro-aliased call (set_pre_exit_functions is #defined straight to SetProperty(string, mixed) in secure/include/compat.h) already documented and fixed on ds386's own copy of this identical file; every sibling room in the same directory (hall.lpc, hall2-5.lpc) already has the call commented out, confirming this never worked under any driver. Commented out identically, matching ds386's own fix verbatim (same file, same lineage).
• lib/interactive.lpc's quit()-path and lib/nmsh.lpc's GetPrompt() both still had the pre-fix ed_cmd()/ query_ed_mode() calls from AGENTS.md §6.2's OLD_ED class (their content had diverged from ds386's baseline enough that the bulk port skipped them) -- applied the identical remedy ds386 already proved: dropped the belt-and-suspenders manual ed_cmd() save/close (the driver's own destruct_object() already flushes a live ed_buffer on destruct under OLD_ED) and dropped the now-unreachable query_ed_mode()-gated custom prompt branch (the driver prints its own ed-mode prompt directly from C while an editor session is open, bypassing this apply entirely).
• 24 domains/<Continent>/virtual/{server,void}.lpc files (one pair per bundled continent-stub domain: Africa, Americas, Atlantis, Australia, AustraliaOutback, China, Europe, Japan, Mexico, Russia, SouthAmericas) all had #define __DIR__ "/domains/<Continent>/ virtual/" -- __DIR__ is a live compiler predefine on this driver (same class as AGENTS.md §7.39), so this is error: Illegal to redefine a predefined value. on every one. This is the exact same virtual-server/void template ds386 hit on its own amigara/ learning domains, just replicated 12× here (once per continent stub) instead of ds386's 1×. Deleted all 24 #define lines (the driver's live __DIR__ produces the identical value, trailing slash included). The two matching top-level scaffold templates (std/virtual_void.txt, std/server.txt, with a literal CHANGEME placeholder) were correctly left untouched -- not meant to compile standalone, matching AGENTS.md's area_room.lpc/open/ prog.lpc template-scaffold precedent.
• Admin account seeding -- see the new AGENTS.md §1.5 Dead-Souls- lineage sub-entry for the full writeup. Summary: this repo arrived PAST its own first-boot install wizard (three real wizard accounts already exist), so the usual "register through the bootstrap wizard" path doesn't apply. Added fluffos to every line of secure/cfg/ groups.cfg (alongside the pre-existing gravity:thurtea:shrub, not replacing them), registered fluffos through the ordinary NEW-player flow (password Mud@2026), then with the driver stopped moved secure/save/players/f/fluffos.o to secure/save/creators/f/ fluffos.o unchanged -- exactly what the stock encre admin command does under the hood (creatorp() checks the live player object's file_name() against DIR_CRES, not any ACL/rank table). Verified live: fluffos immediately shows the admin room-path annotation and banner, and goto/wizard commands work with no further seeding.

4. SEVERE FINDING: domains/omega, domains/common, and the bundled domains/std example area are ALL non-functional -- a whole missing base-object framework, never committed to this repo

This is the single most important finding of this onboarding. See AGENTS.md §7.164 for the full catalogued writeup; summary here.

The task brief for this onboarding (and the repo's own file layout) strongly suggested domains/omega -- 150 files describing "Omega Station," a detailed multi-floor sci-fi space station with NPCs, a shop, and a working elevator -- was this lib's standout content. It is not functional. Every one of the 150 files (148 with real code, 2 are empty 0-byte stubs) fails to compile, because they inherit classes that are #defined nowhere in the entire repository: LIVING, M_VENDOR, CONTAINER, M_OPENABLE, M_WEARABLE, M_GETTABLE, ARMOUR, INDOOR_ROOM, ELEVATOR. Checked exhaustively: the whole mudlib tree, the bundled fluffos-2.23-ds03/ driver source, and even the non-mudlib top-level extra/wolfpaw/ directory (which turned out to be classic MudOS driver-BUILD headers -- macros.h/configure.h -- not mudlib code at all, a red herring this session chased down before ruling it out).

domains/common (33 files, shared sci-fi item/ship infrastructure -- WEAPON, DRINK, ITEM_DECAY_FOOD, SPACESHIP) hits the identical problem, evidently meant to back domains/omega. And bundled stock Dead Souls 3.9 itself ships a large Lima-style "builder school" tutorial/example area at domains/std (315 files -- a completely different missing-macro set: KEY, OBJ, PARAM_OBJ, BOOK, VEHICLE, ADVERSARY, STOCK_MASTER, RANGED_WEAPON, LIVING, INDOOR_ROOM) -- same failure shape, different intended framework (this one looks like it's meant to be a well-known classic LPMud "Lima"/LPUniversity teaching area, judging by the school/A/F/M/O/R/ W/X letter-subdirectory convention and a lima_guide.lpc/domains/ std/lima/ presence).

Together these three trees account for 496 of the lib's 519 total lpcc_check.sh compile failures (148 + 33 + 315) -- one root cause, not 496 separate bugs.

This is not a gap this project's conversion introduced. secure/ include/compat.h (gated live: COMPAT_MODE 1 in global.h, genuinely wired into every object's global include) proves the repo's own author was ACTIVELY building a compatibility shim for exactly this purpose -- it correctly maps ROOM/OBJECT/DAEMON/MONSTER/VAULT and ~50 lowercase set_*/query_* verb names (set_short→SetShort, etc.) onto this lib's own Dead Souls CamelCase API -- but it simply never got the additional ~18 class macros these three content trees actually need. The upstream repo's own git log (only 10 commits total: "Initial commit" bulk-imports these large trees, then all follow-up commits are small numeric tweaks -- "money pops up a lot more than I thought", "multiple fixes credits added mana now ppe") is consistent with one developer's personal, mid-construction mud: the big content trees were bulk-imported once and the compat layer was never finished before development moved on to other systems.

Confirmed live, not just via the batch compile sweep. Logged in as the seeded fluffos admin and ran goto /domains/omega/room/ elevator_center -- the driver echoed the exact compile error interactively (error: syntax error, unexpected L_IDENTIFIER ... inherit ELEVATOR;) followed by *No program in object '/domains/ omega/room/elevator_center'! on the subsequent reference, then Could not load that location. goto /domains/omega/npc/hugo failed the same way. As a control, goto /domains/Praxis/supply2 (stock Dead Souls content, fixed above) loaded correctly and printed its real room description -- confirming the failure is specific to the three missing-framework trees, not a general problem with goto/this driver/this lib's boot.

No attempt was made to author the missing framework. Guessing at correct implementations for LIVING/CONTAINER/ARMOUR/ELEVATOR/ M_VENDOR/etc. (and the completely separate KEY/OBJ/ADVERSARY/ STOCK_MASTER/etc. set for domains/std) from ~500 call sites' worth of usage would mean writing a second mudlib engine from scratch, not fixing a bug -- this project's established policy (ds386's own LIB_CAPTURE/jar.lpc precedent: "implementing an entire mechanic from a one-line hint would be inventing content") applies at much larger scale here. domains/omega and domains/common are left exactly as shipped (uncompilable, documented); nothing was deleted.

What DOES actually work: the stock Dead Souls 3.9 base itself, and every domain built on its real, existing API -- Africa/Americas/Atlantis/Australia/AustraliaOutback/China/ Europe/Japan/Mexico/Russia/SouthAmericas (thin stock-base stubs, as expected and documented for Dead-Souls-lineage variants elsewhere in this corpus), campus/town/Praxis/Ylsrim/default/ amigara (the standard Dead Souls demo areas, same as ds386), plus this lib's own Rifts-flavored additions layered onto the REAL API: Rifts races (greathorneddragon alongside the stock dwarf/elf/human/ etc. set, secure/cfg/races/greathorneddragon), a P.P.E. (Potential Psychic Energy) resource shown on the prompt and score sheet alongside HP/SP, and Rifts-flavored currency/vendor content in the town.

5. lpcc_check.sh compile sweep

Ran under (ulimit -v 8388608; ...) given the tree size (3219 files, larger than ds386's 2427). Final: 2700/3219 pass, 519 fail, all 519 fully triaged and explained:

• 496 -- the missing-framework finding above (§4): 148 domains/omega, 315 domains/std, 33 domains/common.
• 9 -- personal wizard-realm scratch/test content (realms/thurtea/tmp/*.lpc ×5, realms/thurtea/legwand.lpc -- a SetPosition(int vs string) type error in a private test item --, realms/gravity/area/room/{startroom,wayne_manor_backroad}.lpc -- missing RACE_D/ROOM_NEWBIE macros and genuine syntax errors --, realms/gravity/tmp/*.lpc ×1). These are individual wizards' own private home-directory scratch/test content, not part of the shipped game world any ordinary player reaches -- left undocumented-further and unfixed, matching this project's treatment of admin-only/ out-of-scope tooling elsewhere.
• 14 -- exactly matches ds386's own already-documented "known non-functional" class (same files, several literally shared byte- for-byte with ds386): obj/stargate.lpc (deliberate author anti-load trap), obj/area_room.lpc + open/prog.lpc (builder-tool template scaffolding, $S-placeholder/relative-include patterns never meant to compile standalone), domains/Ylsrim/broken/jar.lpc (pre-existing incomplete content in the archive's own broken/ bin), secure/cmds/admins/opcprof.lpc (calls a MudOS efun never ported to FluffOS), secure/tmp/thurtea_CMD_EVAL_TMP_FILE.lpc (an auto-generated cmd eval scratch file), secure/daemon/i3router/ server.lpc (optional admin-only Intermud-3 tooling, undefined ROUTER_NAME -- not on the boot/login/play path, matches ds386's own explicitly-left-alone i3router/imc2/ftp admin-tool list), and daemon/help.lpc + secure/daemon/mudinfo.lpc + domains/campus/{chamber/elevator,room/monty,room/plab2}.lpc + domains/campus/npc/{charles,charly}.lpc (zero actual "error:" text in the failure log -- the AGENTS.md §10.4 "isolated single-file lpcc-batch-mode artifact" class; daemon/help.lpc in particular is proven fine live -- the driver's real boot compiled it clean and help-adjacent features worked throughout the playthrough).

Zero unexplained failures.

6. Live playthrough (native driver)

Full continuous session via scripts/mudclient.py against ~/src/fluffos/build-debug/src/driver config.fluffos.

1. Fresh boot: log/debug.log clean, zero uncaught errors (only the expected benign nosave-on-function warnings, AGENTS.md §4.3), Accepting telnet connections/Initializations complete. 2. Registered a fresh test character, Qintestomegab, through the full flow (name → confirm → age gate → screen-reader prompt → password/confirm → gender → email → race pick human → two news pagers) -- landed in the start room wearing a t-shirt and jeans, the HUD prompt showing hp/ppe/sp (P.P.E. being this lib's Rifts- specific addition to the stock HP/SP pair). 3. look: correct room description/exits every time, across multiple rooms (start room, Horace's General Supply Shop, First Village Bank with "Zoe the bank teller" present). 4. score: correct full character sheet (name, level, race, class, native town, faith, birth date, training points, feeling/food/ drink/alcohol status, XP-to-next-level, customization points) -- confirmed both as a fresh player and as the seeded admin. 5. i (inventory): correct (starting Player's Handbook + worn t-shirt and jeans). 6. Reconnect path verified distinct from fresh registration (a second connection while the first session was still open correctly showed Reconnected./[X has rejoined]). 7. Admin verification: seeded fluffos per §3 above; confirmed live via the admin-only room-path annotation on look, the admin-FAQ welcome banner, and successful goto into real (stock) content. 8. Omega domain: directly navigated via goto (see §4) -- confirmed non-functional live, exactly matching the compile-sweep finding. This was the task's own explicit ask ("navigate into the omega domain specifically to verify its content actually works") and the honest answer is that it does not, for reasons entirely external to anything this onboarding pass could fix. 9. quit: clean disconnect (Please come back another time!, [Fluffos quits Rifts DS], worn items correctly stripped on the way out).

7. WASM status

Measured 2026-09-03 against the shared ~/src/fluffos/build-wasm. secure/sefun/sockets.lpc was already stubbed, but the eager secure/sefun/sefun.lpc still called efun::socket_address() unconditionally -- "Unknown efun: socket_address", simul_efun not loadable, boot failed. Guarded with #ifdef __PACKAGE_SOCKETS__ (same as dsIII/ds386). After that, scripts/wasm_client.js logged in as fluffos / Mud@2026 into First Village Bank, score showed "Fluffos the unaccomplished", hp: 440/440, quit printed "Please come back another time!" secure/daemon/instances.lpc still fails to compile on socket_status() during quit's channel broadcast — a graceful skip, not on the login path. Shop/combat/death were not exercised this pass. Do not loop-reboot (I3).

§10.7 deep functional test (2026-08-31, round two)

Full continuous playthrough against the shared ~/src/fluffos/build- debug/src/driver config.fluffos (no dedicated worktree needed), deliberately exercising the WORKING stock-Dead-Souls content per this session's own brief -- the confirmed-broken domains/omega/domains/ common/domains/std gap (§4 above, AGENTS.md §7.164) was not re-litigated or re-poked at beyond a single deliberate negative-control clone (below). A fresh throwaway mortal registration (Vantharion, human) through the full chargen ritual (name validation, age gate, screen-reader prompt, password, gender, email, race pick) landed correctly in the start room; enter town reached the real, working village hub. Also re-verified the seeded fluffos admin account (§1.5/AGENTS.md's Dead-Souls admin-seeding note) still logs in correctly and shows the wizard room-path annotation.

What was tested and confirmed working -- zero programming bugs found this pass

• Movement/navigation: start room -> town intersection -> Saquivor Road (a talking beggar NPC gave a real map item) -> the Healer's Guild building, all with correct descriptions/exits/NPCs every time.
• Shop/economy: the Healer's Guild's real buy-a-service shop (buy claritin from james) correctly refused an empty-walleted character with "You don't have enough universal credits to buy that." -- a genuine, live-verified insufficient-funds economy check, not just a list read.
• Real combat, extensively: cloned a real stock-content NPC (/domains/town/npc/orc, the ordinary Dead Souls API -- inherit LIB_NPC) as the admin and fought it to the admin's own death. Limb-targeted damage was exchanged realistically in both directions for well over 30 rounds (hits, misses, "flourish of attacks," stunning blows, per-limb messages) with zero crash and zero debug.log growth throughout.
• Death and full resurrection cycle, live end-to-end -- normally one of the two hardest-to-reach §10.7 checklist items, verified for real here because the admin-vs-orc fight above went all the way: SYSTEM <death> Orc has slain Fluffos. -> landed in /domains/ default/room/death (YOU ARE DEAD!, "You are in formless void without substance and outside time.") -> the room's own regenerate command worked correctly, rematerializing the character back at /domains/default/room/start with restored (if reduced) HP/PPE/SP. Not gated behind wizardp() the way some other libs' ghost-recovery flows are (AGENTS.md §7.68/§10.7 item 6a) -- this lib's own death system handles a wizard character exactly like an ordinary one.
• quit / reconnect: clean disconnect (Please come back another time!), log/debug.log never created at any point in the whole session (before, during, or after the death/regen sequence). A fresh reconnect as fluffos correctly restored the post-respawn location (/domains/default/room/start) and HP/PPE/SP state from the save file.
• Negative control, confirming §4's finding is still accurate and scoped correctly: deliberately tried clone /domains/std/monster/ troll (part of the already-documented broken domains/std tree) -- got the exact same inherit ADVERSARY; compile error §4 already catalogued, then successfully cloned the WORKING /domains/town/npc/ orc instead for the real combat test above. Confirms the broken/ working boundary from onboarding is still exactly where it was documented, not encroaching on stock content.
• ~220s (3+ hour, see below) idle long-sit boot watch: debug.log never created; stdout showed zero new errors beyond the one expected domains/std/monster/troll.lpc compile error from this session's own deliberate negative-control clone above.

Minor, unresolved cosmetic observations (not fixed -- neither blocks anything or has a confirmed severe effect; scope discipline against guessing)

• HP/PPE/SP occasionally render as a raw float (hp: 132.000000/440 instead of hp: 132/440) after the death/regenerate sequence -- some stat-regen calculation along that path isn't truncated back to an int before being substituted into the prompt template. Purely cosmetic (regeneration and the death/respawn cycle both worked correctly regardless); not root-caused given time constraints.
• domains/Praxis/supply2.lpc's add_sky_event((: "shop_closing" :)) triggers this driver's "Function pointer returning string constant is NOT a function call" warning at compile time -- the same MudOS single-string functional-literal shorthand class already catalogued in AGENTS.md §8.23 (confirmed there to silently never fire on this driver). Here it would only affect a shop's scheduled "closing" cosmetic flavor event, not blocking list/buy/sell themselves; not chased further given the narrow, cosmetic blast radius and time budget.

Not reached this session, flagged honestly

• Skill/guild-class progression (as distinct from the Healer's Guild's plain buy-a-service building, which is a shop not a class/skill guild): not exercised -- time was budgeted toward the admin combat/death/respawn sequence instead, which this project's own methodology explicitly treats as one of the two hardest-to-reach checklist items and therefore worth prioritizing when it becomes reachable.
• A full two-sided paid shop transaction (buy AND sell, with correct change): only the insufficient-funds refusal path was reached; a fresh mortal character never accumulated enough starting currency to complete a real purchase in the time available.

An extended idle window (not a soak-test bug)

The driver was left running considerably longer than the standard ~200s idle target this pass (a background-task notification delay outside this session's control stretched the wait to several hours of real wall-clock time) -- reported here for the record since it's an unusually long uptime, not because anything was found: debug.log remained completely absent the entire time, and memory growth (~1.5GB -> ~5.6GB RSS) is consistent with this driver's own already-documented ambient NPC heart_beat()/lazy-compile behavior (AGENTS.md §10.8) over that much extra wall-clock time, not a new leak specific to this lib.

Files modified this pass

None -- this pass found zero programming bugs. Runtime state generated during testing (the fluffos admin's updated save, the throwaway Vantharion character's save/mailbox, and a few daemon .o save files) was reverted before committing, per this project's own runtime-state policy.

§10.7 deep functional test (2026-09-01, round three) -- guild/class progression was completely broken, fixed

This pass deliberately targeted the two angles round two's own "Not reached" section flagged: skill/guild-class progression, and a completed two-sided (buy+sell) shop transaction. It also tried tell (including a nonexistent recipient) and a battery of malformed/empty command inputs. Both flagged gaps turned into real findings.

BUG 1 (SEVERE): every Praxis guild-join point silently failed to

ever change the player's class -- SetClass() called where

ChangeClass() was required

Symptom: domains/Praxis/fighter_join.lpc's become fighter (and the identical pattern in cleric_join.lpc, mage_join.lpc, monk_join.lpc, kataan_join.lpc, rogue_join.lpc) always printed its success message ("The Great Warrior initiates you into the class of fighters.") but score/skills afterward still showed the character as Explorer with none of the class's real skills -- 100% reproducible, on a fresh level-1 character, every time, for every one of the six guild-join rooms.

Root cause: all six become()/join() handlers call this_player()->SetClass(class_name) directly. lib/classes.lpc's SetClass() is NOT the entry point for a fresh character's *first* class -- its own header comment says so ("pass by reference... please see the FAQ") and its logic proves it: if(Class){ if(!high_mortalp()) return Class; ... } -- if the player already has ANY non-zero Class value (and every character does, since lib/player.lpc:317 sets Class = "explorer" for everyone at creation), SetClass() treats the call as multi-classing (bolting a second class onto an existing one) and silently refuses unless the caller is high_mortalp() (a non-creator player above level 24 -- Rifts DS's own "veteran mortal" tier for genuine multi-classing, secure/sefun/ pointers.lpc:95). A fresh level-1 (or even a creator, since high_mortalp() explicitly excludes creatorp() accounts too -- confirmed live: fluffos itself could never pass this gate either) explorer can never satisfy that gate, so the assignment is dropped on the floor every time, with the success message printed regardless because it's written to the screen *before* the doomed SetClass() call.

The correct entry point for a fresh explorer's first class is ChangeClass() (same file, lib/classes.lpc): it explicitly does Class = 0 before delegating to SetClass(), which is exactly what bypasses the multi-class gate for a legitimate first-time assignment, then restores the player's level via ChangeLevel(lvl). Proof this is the established, working convention in this very lib: domains/town/npc/herkimer.lpc:111 (the real, working, already round-two-tested Mage's Guild join command, "ask herkimer to join") calls ob->ChangeClass("mage"), not SetClass(). The six Praxis files are the only call sites in the whole corpus using the wrong one (grep -rln ChangeClass turns up exactly lib/classes.lpc, herkimer.lpc, Ylsrim/npc/roshd.lpc, and secure/lib/connect.lpc -- none of the six broken join files).

Fix: this_player()->SetClass("X") -> this_player()-> ChangeClass("X") in:

• domains/Praxis/fighter_join.lpc:78
• domains/Praxis/cleric_join.lpc:67
• domains/Praxis/mage_join.lpc:77
• domains/Praxis/monk_join.lpc:66
• domains/Praxis/kataan_join.lpc:66
• domains/Praxis/rogue_join.lpc:73

Live-verified: fresh boot, admin fluffos (level 1, Explorer) -> goto /domains/Praxis/fighter_join -> become fighter -> score now correctly shows Class: [Fighter] -> skills now correctly shows the real fighter primary skill set (blade attack, blunt attack, melee attack, multi-hand, multi-weapon, etc., all at the class file's configured 4/8 starting level) in place of Explorer's skills. Before the fix, score stayed Explorer and skills stayed unchanged no matter how many times become fighter was run. debug.log stayed empty and stdout showed no new errors across the whole test (expected -- this was a silent no-op, not a crash, per AGENTS.md §10.9's "invisible failure" class).

Known residual content gap, NOT fixed (out of scope, same shape as the already-documented domains/omega/common/std framework gap but much smaller): secure/cfg/classes/ only ships cleric, explorer, fighter, mage, and thief class definitions -- monk_join.lpc, kataan_join.lpc, and rogue_join.lpc call ChangeClass("monk")/("kataan")/("rogue"), none of which have a matching config file. Even with the SetClass->ChangeClass fix applied, CLASSES_D->SetClass() (daemon/classes.lpc:123) returns immediately for an unconfigured class name, so become monk/become kataan/become rogue still silently leave the player as Explorer -- this is a missing-content-data gap (nobody ever authored a monk/ kataan/rogue class file), not a wrong-function-call bug, and authoring three balanced class skill-lists from scratch would be inventing content, not fixing one -- left alone per this project's established policy. fighter/cleric/mage (and presumably thief, not separately live-tested but config-file-complete and structurally identical) are now fully functional via their Praxis join rooms.

BUG 2 (moderate, fixed as part of the same investigation):

query_cap_name()/query_name()/query_gender() are undefined

on the player object -- three missing entries in the lib's own

compat shim

While root-causing BUG 1, call me->query_cap_name() (admin call command, a direct function_exists() probe) confirmed this function plain does not exist anywhere on the player object -- grep -rln "->query_cap_name(" domains/Praxis turns up 51 call sites across 30 files (plus one in realms/gravity/cmds/testooc.lpc), all from the same bulk-imported content lineage documented in §4 above. Two siblings, query_name() (40 call sites) and query_gender() (8 call sites), are equally undefined.

This is the exact same class of gap as the already-documented secure/include/compat.h shim (§4: "the repo's own author was ACTIVELY building a compatibility shim... but it simply never got the additional ~18 class macros" for the three fully-broken trees) -- except this time the shim already has 50 other query_*/set_* mappings covering exactly this kind of naming-convention mismatch (#define query_class GetClass, #define query_open GetOpen, etc.), it just never got these three specific ones added, even though the Praxis tree that needs them is otherwise fully working, non-broken content. Confirmed live: before the fix, preview()'s say(this_player ()->query_cap_name()+" seeks to learn about fighter.", ...) line never produced its "X seeks to learn..." room broadcast to bystanders (silently absorbed); query_class() -- which DOES have a compat.h mapping already -- correctly resolved and printed "Welcome, explorer!" the whole time, proving the macro-substitution mechanism itself works fine and is exactly what was missing for the other three names.

Fix: added three lines to secure/include/compat.h, following the file's own established pattern exactly:

#define query_name                      GetName
#define query_cap_name                  GetCapName
#define query_gender                    GetGender

GetName(), GetCapName(), and GetGender() are all real, extensively-used lfuns already defined on the player object (lib/ player.lpc:531,563, lib/race.lpc:216) -- confirmed via grep -rln GetCapName returning 60+ genuine call sites across the working parts of this same lib (verbs/players/bump.lpc, secure/cmds/players/ tell.lpc, secure/daemon/chat.lpc, etc.), so this is restoring the established convention, not inventing a new one. Two stray hits inside already-broken domains/std/domains/common files (which don't compile at all, per §4) are harmless dead code either way. Because global.h (which conditionally #includes compat.h under COMPAT_MODE) is this driver's configured global include file, every .lpc file in the mudlib picks up the three new mappings automatically on next compile -- no per-file edits needed beyond the one header.

Live-verified: post-fix, call me->query_cap_name() (raw function_exists() probe) still correctly reports "not in OBJ" (a runtime string handed to call_other is never macro-expanded -- expected, since #define is a compile-time textual substitution over literal source identifiers, not a runtime symbol lookup), but the *compiled* become()/preview() functions in fighter_join.lpc -- which contain the literal identifier query_cap_name in their own source, and get recompiled fresh against the updated compat.h on driver restart -- now correctly resolve to GetCapName() at compile time. No regressions: full driver reboot, zero new compile errors, zero debug.log/stdout errors across the whole remaining test session.

Skill training (organic teacher-NPC route), verified working

With fluffos reverted to Explorer (which carries magic attack/ magic defense as innate secondary skills) and 5 training points granted via call me->AddTrainingPoints(5) (the admin-command shortcut this project's own brief for this pass explicitly sanctions in place of a lengthy real XP grind), goto /domains/town/room/ training -> ask radagast to train magic attack correctly matched against lib/base_trainer.lpc's eventTrain() (found the skill in the player's own skill list, found available training points, started a real multi-stage ContinueTraining() call_out sequence) and printed "Radagast begins teaching you about the skill of magic attack." -- confirming the organic NPC-trainer skill-progression path (distinct from both the Healer's Guild buy-a-service shop and the Praxis guild-join mechanism above) is genuinely wired up and functional. Not chased to full completion (the training sequence runs across several TRAINING_WAIT-spaced call_out()s) since the mechanism's entry path was the object of the test, not the multi-minute wait.

Two-sided shop transaction, completed and verified end-to-end

Registered a fresh throwaway mortal (Testshopper, human), granted currency via call %testshopper->AddCurrency("universal credits", N) (admin shortcut, same sanctioned pattern as above -- note the lowercase key name: find_player()/to_object()'s %name token lookup is case-sensitive on the saved keyname, %Testshopper capitalized fails with "Cannot identify any object" while %testshopper succeeds). At the Healer's Guild (domains/town/room/ healer, reached via enter town -> north -> north -> enter guild):

• Buy: buy healing slip from james with 29000 universal credits on hand correctly charged 20000 ("James exclaims... Here is a healing slip for 20000 universal credits!"), left exactly 9000, and added "A healing slip" to inventory.
• Sell: sell slip to james correctly paid back 10000 (half of the 20000 buy price, per lib/std/vendor.lpc:eventBuy()'s own cost = cost / 2 markdown logic for player->vendor sales) for a final balance of 19000, and correctly removed "A healing slip" from inventory.
• Negative control: sell pill to james (a claritin bought earlier) was correctly refused ("I do not buy things like a claritin pill.") with zero currency/inventory side effects -- claritin.lpc never overrides SetVendorType() so it stays the lib/meal.lpc default VT_DRINK, which doesn't match James's own SetVendorType(VT_HERB). This is correct, intentional GetVendorType() & GetVendorType() bitmask economy logic (a healer buys herbs, not the medicine he dispenses), not a bug.

Both the earlier round-two insufficient-funds refusal and this pass's fully-funded buy+sell round trip are now live-verified for the same shop, closing out this checklist item completely.

tell, briefly

tell testshopper hello there (creator -> mortal, both connected) worked correctly in both directions and correctly charged 15 PPE (a deliberate Rifts-flavored "telepathic tell costs psychic energy" design choice, not a bug). tell <nonexistent-name> hello fell through to a confusing but 30-year-old stock secure/cmds/players/ tell.lpc (Descartes of Borg 950523) code path that attempts an inter-mud-instance tell lookup and then prints the generic "Tell whom what?" instead of a clear "no such player" -- odd UX, but it's unmodified stock Dead Souls library logic (not this lib's own bug), causes no crash, and is exactly the kind of pre-existing-library "design as shipped" case this project's scope explicitly says to leave alone.

Adversarial/malformed input, briefly

become (no arg), become fighter fighter fighter, become ''''"";;--, buy (no args), buy 99999999999999999999 from james, sell (no args) all produced sane rejection messages ("Become what?", "You cannot become that here.", "Buy what from whom?", "There is no james here." -- correctly, since james wasn't in the room this was tried from -- "Sell what to whom?") with zero crashes and zero debug.log/stdout growth.

Files modified this pass

• libs/riftsds/work/secure/include/compat.h -- added query_name/ query_cap_name/query_gender -> GetName/GetCapName/ GetGender mappings (BUG 2 fix).
• libs/riftsds/work/domains/Praxis/fighter_join.lpc, cleric_join.lpc, mage_join.lpc, monk_join.lpc, kataan_join.lpc, rogue_join.lpc -- SetClass() -> ChangeClass() (BUG 1 fix).

Runtime state generated during testing (fluffos's class reverted back to Explorer via call me->ChangeClass("explorer") before finishing, the throwaway Testshopper character's save/postal files, and the incidental daemon/RELEASE_NOTES_HTTP/mudlist.txt .o/txt churn from having the driver up) was reverted before committing, per this project's own runtime-state policy.

Heartbeat hobbled(this_player()) (2026-09-04)

Two-sided shop already verified 2026-09-01; not re-run. Same lib/body.lpc heart_beat hobbled(this_object()) plus disable.lpc objectp guard as dsIII. Verb-file checks left alone. Admin fluffos / Mud@2026 logged in after the edit (village bank / Zoe). Save-file churn from this boot left uncommitted.

2026-10-03 — elevator buttons

domains/town/obj/ebutton1.lpc and ebutton2.lpc called car->SetDoor(1), the name the elevator had before it renamed its own function SetDoorClosed (it shadowed LIB_EXITS::SetDoor(dir, file)); the old name resolves to the inherited exit function and the call is a silent no-op, so the call buttons never closed the car door. Changed to SetDoorClosed(1), as in dsIII and ds386 (KB 06 §7.211). The rest of this lib's compile warnings are still to do (queue: Dead Souls siblings).

深度功能测试（§10.7，2026-10-03）— compile warnings and inherit diamonds

Method: as in dshakkard (same twin of ds386, KB 06 §7.213-§7.216): ds386's commit applied with patch (1476 files, 42 hunks and 31 absent files left over), scripts/lpc_warnings.py riftsds --fix for the mechanical hunks that did not apply (183 nosave functions -> protected, 23 dropped, 27 escapes, 8 varargs, 2 bare returns, 19 unused locals), scripts/lpc_move_decl.py (master.lpc), scripts/lpc_fix_strptr.py (orc_valley/guard.lpc), scripts/lpc_fix_unused_init.py (cmds/players/equipment.lpc), scripts/lpc_diamonds.py --fix-redundant (domains/town/meals/worm.lpc: LIB_BAIT_WITH already inside LIB_MEAL, which this lib has and ds386 does not) and by hand the empty if in daemon/intermud.lpc, room.lpc's SetSky() and opcprof.lpc (the efun is gone from FluffOS). Oracle (263 programs, HEAD against working tree): 1763 public -> protected flips, 46 definer changes (the same explicit choices as dshakkard, including the horse's five mount applies whose bodies are identical in LIB_MOUNT and LIB_LIVING), 455 duplicate variables removed or renamed, 92 enter/knock/ scratch/door functions the three dummy hybrids (trashcan, riverwater, seawater) no longer inherit, 88 glow rows. Live, paired: the trainer (HEAD "I am already training you!" after a pupil left and came back within two seconds, now "You will have to start your studies anew"), gallows and stone (HEAD printed long_func), fixtures, search, zpem, horse, death and corpse.

Result: 6213 -> 534 distinct diagnostics, none of them a warning in a loadable file. 533 are the errors of the 518 files this lib has never been able to compile (the Lima-style domains/std, domains/omega, domains/common and creator-realm content, triaged in the 2026-08 sections above: missing ITEM_DECAY_FOOD/ GUILD_D/RACE_D/ROOM_NEWBIE and headers that do not exist); the last is an Expression has no side effects in a creator's eval scratch file the archive shipped (secure/tmp/thurtea_CMD_EVAL_TMP_FILE.lpc, which the eval command overwrites on every use). The comma pointers (: this_object(), "fn" :) of the Praxis domain (29 sites in 21 files) are still to do (KB 04 §6.10).

深度功能测试（§10.7，2026-10-04）— Praxis comma-form function pointers

The 30 comma-form pointers in the 21 Praxis files and the target fixes made in ds386 (see its NOTES section of the same date: SetSearch handlers return their text, the helm asks GetMorality(), test_invis takes the viewer, look_at_vault/look_at_manhole restored as text) carry over here: scripts/lpc_twin_port.py adopt ds386 riftsds --pre HEAD --apply (all 21 are token twins: 12 byte-identical, 9 spliced into riftsds's own spacing). All 21 files compile without a diagnostic, and a live spot check (search field, search mounds, look at mound/manhole/vault/treats, the planning room, a voting hall) answers as in ds386.

深度功能测试（§10.7，2026-10-04）— last #pragma no_warnings removed

domains/std/{Attic,Church,wiz_hall}.lpc still opened with #pragma no_warnings. They inherit /std/indoor_room, which this lib does not have, so they never compiled and the pragma hid nothing; it is gone, and no lib file carries it any more.

深度功能测试（§10.7，2026-10-04）— paths named in another case

The archive comes from a case-insensitive host: files are stored as BAGUA0.lpc, longjing.C, d/SHAOLIN/, the code names them in lower case (__DIR__ "bagua0", "tea/longjing", "/d/shaolin/..."), so the rooms, NPCs and items behind them could not be loaded (a .C file is not an .lpc file; KB 03 §4.5). scripts/lpc_case_paths.py --convert --apply riftsds: 2 file(s) renamed to lower case (.C -> .lpc, .H -> .h) and 1 path reference(s) in 1 file(s) rewritten to the new spelling; 0 skipped, 0 ambiguous. Load check of the 4 objects that were renamed, edited or name a renamed path (HEAD against the working tree, an object is the same under any spelling of its path): 0 PASS -> 1 PASS, 0 that loaded before fail now, 1 more load.

2 files moved (2 .lpc); mostly under domains/std/rooms (1), domains/std/Wizroom.lpc (1). A rescan with scripts/lpc_warnings.py riftsds is still due: the renamed files are compiled for the first time.

深度功能测试（§10.7，2026-10-05）— 主控 author_file()/domain_file()

驱动的 mudlib_stats 包在主控载入时各调用一次 author_file(<主控文件>) 和 domain_file("/")，要求返回字符串。本库这两个 apply 只为 domains / realms 目录下的文件返回名字，其余返回 0，所以每次启动都打印 author_file() in the master file does not work, using root_uid as fallback 和 domain_file() in the master file does not work, using bb_ui as fallback 两行，网页版每位访客的终端里都看得到。secure/daemon/master.lpc 改成与 fluffos 组织维护的 Dead Souls 上游（deadsouls_fluffos）相同的写法：其余文件的域返回 "BACKBONE"，作者返回 "NONAME"。域等于 backbone 域时驱动让对象沿用创建者的域，与返回 0 时相同；作者统计（author_stats()）多出一行 NONAME，汇总 realms 目录以外的对象（KB 06 §7.222）。用 lpcc config.fluffos /nonexistent.lpc 前后对比，主控载入段只少了这两行。

• 2026-10-14: FluffOS 970d9f09 warns Float value truncated to int (#1426). 16 sites in 10 files: assignments, initializers and returns now spell the truncation to_int(...) (same value); x op= <float> converted the float before the operation (zeroing x for a factor below 1) and is now x = to_int(x op ...): /lib/teller.lpc（total +=）, /secure/daemon/players.lpc（seed *=）. Changed objects loaded before and after in fresh drivers, no regressions (scripts/lpc_fix_float_trunc.py, KB 04 §6.10).
• 2026-10-14 (same rescan, by hand): lib/combat.lpc set the default Wimpy = 0.20 into an int, so it was 0 and auto-flee never fired for anyone who had not set a wimpy value. Wimpy is a percentage of max health (wimpy command, SetWimpy(95) on NPCs, compared with percent(hp, max)), so the default is now 20 and SetWimpy/GetWimpy take and return an int. This is the same fix as brassring, and the default the later Dead Souls releases use (ds386, dsIII, fluffos/dead-souls: Wimpy = 20). Gameplay-visible: characters flee at 20% health by default. The 11 changed objects load, with no regressions.

Save files the onboarding transcode broke (2026-10-15, KB 03 §4.4)

/save/intermud.o: the archive's Windows-1252 bytes (a \x92 apostrophe, ¿, ¡) were run through iconv -f gb18030 -c, which dropped an undecodable byte together with the quote after it, so restore_object() stopped with Illegal mapping format and the object started empty. Re-decoded from raw/ as Windows-1252 (scripts/lpc_save_encoding.py --from-raw); the driver restores each one now (scripts/lpc_save_restore_check.py).

Save files restore_object() refused (2026-10-15, KB 03 §4.4)

domains/campus/save/charles.o、domains/campus/save/charly.o、secure/save/mudinfo.o: a raw newline inside a string (save_object() writes it as \r; restore reads line by line and the string never closes) or a blank line (Illegal file format), so the object started empty. Newlines written as \r, blank lines dropped (scripts/lpc_save_encoding.py --from-raw); the driver restores each one now (scripts/lpc_save_restore_check.py).

Restored base-class macros (2026-10-18, KB 04 6.13)

inherit IDENT; named a macro that no header of this lib defines (syntax error, unexpected L_IDENTIFIER, expecting L_STRING or '('), so the object never loaded; the base file is in the lib and other libs map the macro to it. Added to the global header secure/include/global.h: BOOK -> /obj/book (2 file(s)) (scripts/lpc_restore_macros.py). Loaded one by one with lpcc afterwards: 0 of 2 pass, the other 2 stop on another error (domains/common/item/quest_todo.lpc: Undefined function set_proper_name; domains/std/lima_guide.lpc: Undefined function add_id).

x != "a" || x != "b" and x == "a" && x == "b": comparison chains of one value that can never differ (2026-10-20, KB 06 7.282)

if (arg != "wan" || arg != "pill") return notify_fail(...) is true whatever arg is, dir == "south" && dir == "north" is never true: the guard under the test never opens, or never closes. The author meant "neither of these" (&&) or "either of these" (||). != chains joined by || became && (only where the statement under the test is a return, a refusal or a guard clause), == chains joined by && became || (scripts/lpc_fix_cmp_chain.py). What it was: prompt / editor messages were never marked MSG_NOWRAP (msg_class == "prompt" && msg_class == "editor"). Changed: /lib/interface.lpc (1).

if (x = CONST): a comparison written as an assignment (2026-10-20, KB 06 7.290)

if (arg = "button" || arg = "pill"), if (!arg || arg = "") return 0;, else if (check = 0 && tempB != ""): = was typed for ==, so the condition is the constant: the branch is always taken for a non-zero number or any string (even "") and never for 0, and the variable is overwritten; the driver does not warn. Every such = became == (scripts/lpc_fix_cond_assign.py, 9 terms). Changed: /cmds/creators/cryptogram.lpc (1), /cmds/creators/rot.lpc (1), /domains/campus/npc/jennybot.lpc (1), /domains/default/weap/grenade.lpc (1), /domains/default/weap/plasma.lpc (1), /domains/default/weap/rpg.lpc (1), /domains/default/weap/torpedo.lpc (1), /lib/lead.lpc (1), /verbs/builders/reload.lpc (1).

A login password over 255 bytes: crypt() raises (2026-10-20, KB 06 7.294)

The driver's crypt() raises (Error in crypt(), check your salt) for a key over 255 bytes, where the old driver hashed any length. A password of 256 bytes or more typed at the login prompt threw an error in the password check and left the connection at a bare > prompt, so the login never completed (the old-password test of password the same); found by the sweep shape of 300 x. /secure/sefun/sefun.lpc now defines a simulated efun crypt(key, salt) of the same name: only an over-long key is cut (to 255 characters, then character by character until sizeof(to_buffer(key)) <= 255: the limit is in bytes, a CJK character takes three) before it is handed to efun::crypt(); every ordinary password hashes as before and the saved hashes are untouched (scripts/lpc_add_sefun.py crypt).

help <topic> on a plain clone, telnet and setreboot without an argument (2026-10-21, KB 06 7.212, 7.336)

help <topic> loads /daemon/help, whose LoadIndices() adds get_dir() results with +: secure/cmds/common, verbs/spells and verbs/undead are directories the archive ships empty and a git clone does not have (git keeps no empty directory; the site recreates them from scripts/wasm_keep_dirs.txt). Booted without them, the first help <topic> printed A runtime error occurred (Bad type argument to +. Had array and int) and the half-built daemon answered every later topic with "no results". Every directory is now read through DirList() (get_dir(pattern) || ({})); help look, help telnet, help races and help commands answer with the four directories removed (PLAY_RMDIRS). Neither verbs/spells nor verbs/undead was in wasm_keep_dirs.txt for this lib, so the site was affected too. The mortal telnet with no argument raised Bad argument 1 to sscanf() Expected: string Got: 0 (the bare-verb sweep: A runtime error occurred), and the usage line it meant to print came from this_object()->help(), a function no object defines: it tests !args || now and prints GetHelp(); the administrators' setreboot called the same missing help() and went on without a return, it prints its syntax and returns now. Changed: /daemon/help.lpc, /secure/cmds/players/telnet.lpc, /secure/cmds/admins/setreboot.lpc.

Creator and administrator commands with no argument, the A* padding (2026-10-21, KB 06 7.336, 7.338)

The creator / administrator verbs had never been run, only compiled. Swept as the seeded administrator (play_sweep_lib.py SLUG --wiz --account fluffos:Mud@2026 --shapes one), seven of them raised on the bare command: areaclone and cryptogram (sscanf(0, ...)), pause (lower_case(0)), html (read_file(0)), encre (player_save_file() runs before the argument is looked at and raises Bad argument 1 to save_file()), resolve and wget (the daemons RESOLV_D->eventResolve(0) and WGET_D->eventGet(0) run sscanf(0, ...) / strsrch(0, ...)). Each tests the argument first and prints its GetHelp(). The A* sefun's find_path() (the creator command astar) died with (s)printf(): Null pad string specified because it pads the map with sprintf("%N'\27's", ""), a pad string that has no display width in FluffOS; it pads with spaces and replaces them, and astar prints its demo route again. The sweep must skip removeraces (it empties the race table: every NPC created afterwards fails in its heart beat). Changed: /secure/cmds/builders/areaclone.lpc, /cmds/creators/cryptogram.lpc, /secure/cmds/admins/pause.lpc, /cmds/creators/html.lpc, /secure/cmds/admins/encre.lpc, /cmds/creators/resolve.lpc, /secure/cmds/creators/wget.lpc, /secure/sefun/astar.lpc.

add_action() registrations of functions nobody defines, second round: objects behind a macro path or that the static scan cannot settle (2026-10-21, KB 06 §7.349)

The first round's static scan could not read objects that inherit through a macro path (a #define ROOM STD_DIR "room/room" whose body starts with another macro) or an inherit it could not resolve, and skipped them as UNSURE; the second round fixed the resolver, rescanned, and asked function_exists() of those objects in a running driver. An add_action("fn", "verb") whose function is defined neither by the object, nor by its headers, nor by its parents, nor by any object derived from it (scripts/lpc_scan_callout_targets.py, descendants checked transitively, then function_exists() in a running driver, scripts/lpc_verify_handlers.py). The driver walks the actor's sentences from the newest one: when that sentence's function does not exist it raises *Function for verb 'drop' not found. and tries no other handler of the verb (an older dead sentence is passed over silently). Removed 3 registrations in 2 files (scripts/lpc_fix_dead_add_action.py), the verbs are left to their other handlers: /secure/obj/machine.lpc (check_it), /secure/obj/memo.lpc (clean_memo, erase_memo). Most frequent: check_it x1, clean_memo x1, erase_memo x1.

Combat probe with the seeded administrator (2026-10-22): the omega NPCs do not compile

scripts/play_combat_probe.py riftsds --account fluffos:... --clone ... clones each opponent into the administrator's room and attacks it at once. Two files of the omega domain fail to clone: /domains/omega/npc/hugo.lpc and /domains/omega/npc/floor8/doc_williams.lpc start with inherit LIVING; inherit M_VENDOR; (Lima macros), which a Dead Souls base does not define (3:9: syntax error, unexpected L_IDENTIFIER, expecting L_STRING or '('): the domain was written for another lib and its NPCs never loaded here. Not changed (a port, not a repair). The Dead Souls NPCs of the lib fight cleanly.

new(MISSING)->move() / carry_object(MACRO(..))->wear() chains guarded (2026-10-22, KB 06 §7.73 addendum)

A statement new(PATH)->move(...); / carry_object(ARMOR_D("x"))->wear(); whose path names no file of this archive: new() answers 0 and ->move() on it raises Bad argument 1 to EFUN call_other(). In create() the object never loads, in reset() every spawn below the first missing one is skipped on every reset. 2 statement(s) in 2 file(s) (new) are now { object new_ob = new(PATH); if (new_ob) new_ob->move(...); } (scripts/lpc_guard_carry_object.py --new: the path is resolved with the macros the file itself sees, an ambiguous macro or a virtual-object path is left). Load check of the 2 changed files, HEAD vs tree: 2 load at HEAD, 2 load now (0 newly loading, 0 regress).

Pointers (: name :) to functions the object never defines (2026-10-22, KB 06 §7.355)

A (: name :) in an item_desc / inquiry / chat_msg_combat names a function that neither the object, its headers nor its parents have a body for: it compiles, and the use raises *Undefined lfun pointer called: name (look THING / ask NPC about TOPIC fail on the spot, a combat or idle chat pointer fails inside the NPC's chat() and the rest of that heart beat is skipped). scripts/lpc_scan_callout_targets.py --kinds=lfun. 2 more function(s) stay missing (other 2): the prototype is there and the body is lost, or it was never written, and no sibling lib's copy of the file can be copied as it is (it lacks the function, needs helpers this lib does not have, or has another signature); nothing invented: /secure/lib/net/ftp.lpc:eventCmdPass, /secure/lib/net/h_ftpd.lpc:eventCmdPass.

log/ subdirectories nobody created: the logging write_file() / log_file() raises (2026-10-22, KB 06 §7.358)

The code of this lib writes logs into 3 subdirectories of log/: log/harass (lib/interface.lpc), log/imc2_server (secure/daemon/imc2server/server_log.h), log/secure/log (secure/lib/connect.lpc). Neither the archive, a git clone nor the site pack has them (no local boot wrote into them, so gen_keep_dirs.py never recorded them), and write_file() raises *Wrong permissions for opening file /log/dlog/money for append. "No such file or directory" for a missing directory: the function that logs stops on that line and what came after it (the reward, the flag, the next step) never runs. This lib's log_file() is write_file(LOG_DIR + file, text) and raises the same error for a missing directory (measured). log/harass, log/imc2_server, log/secure/log are now in scripts/wasm_keep_dirs.txt (scripts/lpc_keep_log_dirs.py; the site and the pristine tree create the empty directories from it). Proof (scripts/lpcc_eval.py, an eval object under /adm/daemons or similar writes /log/zzdirprobe/..., fresh processes): the measured table in KB 06 §7.358.

Bare wizard command with a 0 argument (2026-10-27, KB 06 §7.336 Also)

After the permission gate, the command still handed 0 to sscanf / read_file / find_living. lpc_fix_bare_arg.py adds if (!arg) (or help(me) / GetHelp()) before the first use. See this commit for the files touched in this lib.