Rage of Mages demo preflight

Source: https://archive.org/details/RageofMages_1020 (item labels it Demo). Download: https://archive.org/download/RageofMages_1020/romdemo.zip 34,080,548 bytes; SHA1 808a0d8a0af3ec5e8b881deee666384cd2f4ed1e. Fetched and hash-verified on temporary bx_n9yzd3f9, 2026-10-12. Private compatibility fixture; no public redistribution/deployment assumed.

ZIP extraction succeeds: RoMDemo.exe and RoMDemo.r00 through .r10. The EXE is a multipart solid RAR SFX, not yet the game executable. 7-Zip23.01 reports Headers Error, Unexpected end of archive, and Unsupported Method for rom.exe. Its output rom.exe is zero bytes and must not be launched or treated as a valid extraction. Source archive is retained unchanged under /home/user/rage-of-mages-demo-preflight/ on the temporary box.

Next try a compatible RAR reader, verify extracted PE/imports and bundled instructions, then run a bounded launch. No launch, gameplay, screenshot, audio/FPS or registration qualification yet. This refills Oni's core lane after its registered browser gameplay reached ops review.

unrar-free0.1.3 rejects the SFX; a separate copy with the63388-byte SFX prefix stripped still fails (Bad RAR file). Ubuntu unrar7.0.7 also reports rom.exe checksum error/Unexpected end of archive. No reader produced a verified payload. Do not call this an emulator crash or conclude corruption solely from these readers. Next inspect volume format/continuation headers and compare an independent demo source or older compatible reader. Exact logs/source identity: scratch/runs/20261012T0320Z-rage-of-mages-preflight.

Volume audit, 2026-10-12 03:33 UTC

Fresh snapshot fork bx_5avf2969 retains the original files unchanged. All 12 volumes have valid RAR signatures (SFX offset 63388, others offset 0). The full-sized continuation volumes vary from 2,915,138 to 2,915,171 bytes, despite matching packed-length headers in r00 through r04. Counting newline bytes gives 8,314–11,951 LF bytes per volume, but zero or one CRLF pair in each continuation volume (two in the SFX). This is strong evidence consistent with destructive text-mode newline conversion of compressed data; the historical transfer mechanism remains unproven. A reader-version change alone is unlikely to recover missing bytes. Do not insert CR before every LF: original bare LF bytes cannot be distinguished from converted CRLF that way.

Reproduction and full counts are audit-volumes.js / volume-audit.json in the same evidence directory. No guest execution or binary repair attempted. The forum's romdemo.zip link points back to this same Archive item and is not an independent source. A separate candidate is GameStar's October 1998 disc: publisher contents https://download.gamestar.de/public/33400/33409/10_1998_cd.pdf lists Demos\\Rom\\Romdemo.exe; Archive item GSCDA1998 contains 1998-10/GS1098.ISO. A bounded ISO9660 HTTP-range reader is checking that directory, without downloading the full 672 MB disc to the persistent box. The directory is confirmed: ROMDEMO.EXE is 42,111,946 bytes at ISO sector 197989, with README.TXT, INFO.TXT and CDV.TXT alongside it. The first large HTTP range timed out after 90 seconds; directory evidence is preserved in disc-fetch.log. This is an independent candidate, not yet a verified payload. The follow-up 1 MB chunked transfer returned HTTP 500. Retained partial bytes are unverified and must not be extracted/launched as a complete installer. Next resume this exact independent source when its HTTP service responds; do not retry the damaged multipart ZIP with further readers. The temporary box was stopped after this source-only investigation; no browser/CLI slot used.

Independent demo recovered and launched (2026-10-12 03:46 UTC)

Direct Archive server https://ia803103.us.archive.org/15/items/GSCDA1998/1998-10/GS1098.ISO accepted 1 MB HTTP ranges for the exact ISO extent identified above. Full installer42,111,946bytes SHA256 c27baf4367b3593e1d0572b7411ffd485319f997075d94ec15344a25a64c83bc. 7-Zip extracts this ZIP SFX successfully (14 files,68,092,880bytes), unlike the damaged multipart source. The PE-format warning is expected for this ZIP SFX; extraction reports Everything is Ok. No installer execution needed. Remote fixture /home/user/rage-of-mages-demo-preflight/independent on bx_qvkda4g8; originals retained. Logs/hashes in0320Z preflight evidence.

Game rom.exe2,059,776bytes SHA256 0efdb79503d9ffc851a08317d8471c300b1afb26b28866c99ef5c9f3da6fe019; image base0x400000, imports DirectDraw, DirectSound and bundled smackw32.dll. All resources are in the same directory. CLI --exe=<dir>/rom.exe --vfs-include=**/* --quiet-api --batch-size=100000 reaches a clear English main menu by3000batches (10.584s). The branch module hash is 9e835f75dfe5d2fcfb91f592641e945f91baadcd20f6751460777f4dba888cd6. Exact commands/logs/screenshots:0347Z-rage-of-mages-first-launch.

Ordinary 3100:click:220:105 opens New Game and reaches the rendered world with opening dialogue by7000batches (0348Z-rage-of-mages-new-game). 7100:click:316:310 advances to the next character's dialogue (0349Z world). These prove menu/dialogue input only, not unit movement or qualified gameplay. The harness warns real ole32/shell32/comctl32 DLL files are absent; imports use available WAT handlers. No ordinal trap occurred in these runs, but full fixture equivalence is not asserted. Browser, registration, touch/audio/FPS remain unverified. No emulator source changes were needed to reach this point.

Ordinary selection and movement verified

0350Z-rage-of-mages-dialogue repeats OK clicks at7100..12100 in1000-batch steps, reaching the world and a top-left tip. Close tip at(290,156) batch14000, select the wizard at(240,309)14500. Selected.png15000 shows her portrait, commands and health/mana bars. Click(240,230)15100 reaches another encounter dialogue; this alone does not expose the unit position.

0352Z-rage-of-mages-movement-followup adds(316,310) clicks at20000,21000, 22000 to progress dialogue. By25000 the selected wizard is visibly near (300,309), displaced from(240,309), while the other character and ground landmarks stay put. These ordinary clicks establish selection and movement; the later clicks can also issue ground orders once dialogue closes, so do not attribute the entire route solely to the15100 click. The15500 intermediate capture is already obscured by dialogue. Final still has a top-left gameplay tip, but both units and the selected character's HUD are clearly visible. Reviewed selected/final images, commands and logs are preserved; no guest state edits. Bounded25000-batch replay exits0. Next local-only registration and browser/touch route, then audio and scene-qualified FPS. CLI batches/s are not FPS; no release approval or public asset distribution is implied.

Local registration and Chromium touch gameplay (2026-10-12 04:05 UTC)

Registration 9bdf532e4 on codex/dr2-registration adds local candidate rage-of-mages-demo, app rage_of_mages_demo, and direct touch with an Esc button/keyboard affordance. It does not add the game to DESKTOP_APPS. The manifest uses explicit manual placement of the verified installer into archive/ROMDEMO.EXE, then existing extractArchive preparation into game. Archive entry-download URLs respond HTTP200 with zero bytes on both data servers; hash checks reject them. No working one-click download is claimed. The full disc SHA1 is cc199d8dc56fc2636843d625a724f9e21467f483 per Archive metadata. Installer SHA1 is 46b0dbb02e6eafa6eaf9c334c635f9744546f094. Prepared14files and game EXE hash match the verified CLI fixture.

20261012T0401Z-rage-of-mages-browser: fresh remote build and Chrome155, same module9e835f75..8cd6. Registered entry reaches menu. Actual Chromium touchscreen taps open New Game, advance all six opening dialogue prompts, close the tip and select the wizard. Step3 has her at screen(257,545); step4 after a tap(350,550) has her visibly displaced near(355,545), while the other character and terrain landmarks remain fixed. This is browser gameplay. Viewport844x390 then shows the entire game with side gutters and visible Esc/ keyboard controls. Step6 after tap(390,260) moves her from about(435,260) to (385,260), confirming touch input still maps correctly at that layout.

Menu/step1..6, exact probe/command JSON, run log, stage and module/browser identity are preserved. Stage complete04:05:44, browser/server finally closed. This is desktop Chromium touch emulation, not Safari or a physical phone. Audio quality, scene-qualified FPS and broader combat/save/load remain open. Core compatibility is ready for ops review; registration is not on main yet.

Integration reconciliation, 2026-10-13

Recovered from origin/codex/wsasocket-capability at 15b37458e during the release merge audit. The local-only registration and its required runtime fixes are now on main. Earlier statements that integration is pending describe the historical run, not the current branch. This recovery adds no fresh gameplay, audio, performance, Safari, or physical-device qualification; retain the recorded run IDs, module hashes, limitations, and failed attempts above.