Skip to content

About

Rust offline editor and reverse-engineering toolkit for the original Variax 600 through PODxt Live; hardware access locked.

Resources

Stars

2 stars

Watchers

0 watching

Forks

Repository files navigation

Variax Workbench for Linux

A Rust replacement under development for the original Line 6 Variax 600, connected by VDI through a PODxt Live. The current milestone is an offline preset editor and reverse-engineering toolkit, with the original guitar and pickup artwork recovered from Workbench 1.75.

Final product target

One Rust application for guitar + POD + USB recording + device maintenance. The first complete rig is Variax 600 through PODxt Live on Linux. Future backends will bring Workbench HD/JTV-69, POD HD500X and other models into the same app, with independently verified capabilities and device-specific implementations. These are roadmap targets, not current support claims. See the product vision.

PODxt Live historically supports USB guitar recording, including processed/clean routing, as documented in Line 6's recording guide. Linux recording, playback and routing remain unimplemented/unverified here.

Experimental software — use entirely at your own risk

This is an independent, reverse-engineered project developed with AI assistance. It is not an official Line 6 product and is not affiliated with, endorsed by, or supported by Line 6, Yamaha, or any equipment manufacturer. No technical support is provided, and no maintenance, recovery service, or continued development is promised.

The software is provided “as is”, without warranty of any kind. Its output, sound, timing, controls and other behavior may differ from the original software. Errors, incomplete reverse engineering, AI-generated mistakes or incompatibilities may cause lost presets or calibration data, unexpected or damaging audio output, unusable devices, or permanent damage to or complete destruction of connected hardware. Tests, backups and the current hardware locks do not guarantee safety or recoverability. Anyone who downloads, modifies, builds or uses this project does so entirely at their own risk and assumes responsibility for that decision.

Contributors and testers are subject to the same risks during development. Use offline tests, saved data and simulations with the equipment disconnected from the computer; keep the USB cable unplugged. Do not connect hardware merely to try a change or because tests pass. Connecting a POD, Variax or other device, including for identification or supposedly read-only testing, is entirely at your own risk. A contribution, test result or project issue does not authorize live device access or establish that a connection is safe.

To the maximum extent permitted by applicable law, the project owner, authors and contributors disclaim all warranties and accept no responsibility or liability for any claim, loss, damage, injury, expense or other consequence arising from this project or its use, whether direct, indirect, incidental or consequential, including equipment damage, loss of data or loss of use, even if they have been advised that such consequences are possible.

Hardware access is locked before device enumeration or opening. The GUI, Rust USB backend and Python probe have no runtime unlock. No preset uploads, flash writes or firmware updates have been performed. Earlier identity/metadata probes produced one current-preset settings capture, not a full guitar backup.

Read the latest generator/firmware results, the full report for changes, validation, risks and remaining work. Use GitHub issues and the project for work tracking; AGENTS.md records the hardware boundary.

Implementation and validation checklist

Keep this checklist current as issues progress. Checked items passed the stated offline checks; they do not certify physical safety or hardware/audio equivalence. The guitar uses piezo sensing; pickup controls are virtual models. Variable coverage distinguishes tested samples from pending ranges and audio behavior. Integration results and remaining risks: offline runtime and HD research.

Detailed evidence and limitations: body interpolation, pickup response, generator results, and backlog.

Verified and passing

  • Rust CLI and offline GUI build, including optimized release builds.
  • Short guided-check offline rehearsal app: five prompts, cancellation/skip handling and protected report saving. Run cargo run --bin variax-check -- plan; see usage and limits.
  • 218 Rust tests per math profile (host configuration and experimental software configuration), Clippy with warnings denied, and formatting checks on Linux.
  • 23 headless GUI tests cover actual dirty-prompt/bank-row clicks, save cancellation/failure/retry, grouped slider drags, operation cancellation/error recovery, artwork recovery, document preservation and offline recording controls/file rejection and POD tone quit protection. Native OS dialogs/window behavior still need manual platform checks; see GUI scope.
  • Shared CLI/GUI edit validation for 34 parameters, 28 body and 32 pickup choices; 84 original-Java endpoint/choice cases and 1,792 body/pickup/slot selections pass. Imported unknown/out-of-range values survive rendering/copying; see domains and limits.
  • Individual preset parsing/editing and protected copy saves: 6 original-Java compatibility checks.
  • L6BA/L6CO browsing, editing, whole-bank undo/redo and protected save-copy in the GUI; three banks/five presets per format compared with original Java, including resized replacement. Bank scope and limits.
  • Offline GUI with recovered artwork, preset controls, undo/redo and guarded document operations; selected headless tests and rendering smoke checks pass.
  • All 43 pickup definitions and 28 body definitions verified against original native serialization.
  • Complete typed Rust body decoding and 748 bit-exact pickup-normalization comparisons.
  • Rust body interpolation routing: 600 native routing-plan comparisons.
  • Actual Rust body-field interpolation: 252 vendor-body outputs + 84 synthetic outputs match every serialized field in debug and release under the documented host-libm oracle.
  • Initialized original-library reference harness: 26 startup routines, 52 complex constants, 110 generator cases and four JNI mapping comparisons pass with disclosed host-libm adapters. This is reference execution, not full Rust or audio parity; scope and evidence.
  • Offline math sensitivity audit: 206,228 distinct math inputs agree between MPFR-256/512 after target-width rounding; glibc differs by up to one ULP on some inputs. MPFR substitution changes four coefficient words in one of 110 packed cases. A separately recompiled Apple source candidate is not an authenticated Mac runtime. See math runtime evidence and limits.
  • Historical Apple 88.3.9 math candidate: 206,228 observed tuples and 110 whole-generator cases complete offline. Both guest and current-CPU x87 candidates match 109/110 recorded glibc outputs and 110/110 existing MPFR candidates. Fresh Linux Rust debug/release portable-profile binaries match all 110 saved Apple candidates; the host profile retains its 110 glibc references. Emulator x87 precision and complete historical-runtime parity remain limited; source, reproduction and boundaries. Default math is unchanged.
  • Local series-case position diagnostic: five distinct neighboring float32 inputs, including four new synthetic settings, retain the same four coefficient-word differences between host/Apple adapters. Six paired executions include a repeated center and checksum checks. Three vendor-free guard tests cover full byte length, checksum-matching truncations and bad checksums. This quantifies a local persistent difference, without locating the internal threshold or resolving historical accuracy; see the diagnostic limits. Issue #5 remains open and #6 remains held for prerequisite disposition.
  • Original embedded Windows math instructions: 206,228 inputs in each of three explicit CPU/precision modes complete offline. Initialized SSE math substitution into the Mac generator matches all 110 MPFR-256 packed candidates. This isolates math effects; it is not full Windows generator/JVM parity.
  • LLVM prerequisite guard: 15 vendor-free dependency tests pass, including genuine indented Arch/Ubuntu output, exact version tokens, duplicate/conflicting-line rejection and portable command mocks with exact argument allowlists. A version-only check accepts the installed reviewed LLVM 23.1.5. This parser correction does not resume the issue #5 scientific investigation or establish math accuracy.
  • Experimental portable-math build: pinned Rust algorithms, 44 centralized math call sites, 110 complete numeric patches, 48 synthetic-resource cases, 828 raw component cases, and 144 complete Java/native preset-file cases match corresponding original-instruction software-math references in debug/release. Software CLI binaries import none of the tested host math functions. The default profile remains host math; profile contract, dependency and limits.
  • Fresh offline research setup: digest-checked archive extraction, isolated hash-locked Python environment, 28-body/43-pickup catalog, eight native slices, derived procedure metadata and an ordered 39-stage reference recipe. Nine synthetic setup guards pass on Linux/Windows/macOS. A fresh uninterrupted replay passed all 39 stages without fixture changes. Replay evidence and retained bounded-run failures are documented; the Linux research environment is separate from application dependencies.
  • Linux offline package builder, explicit digest manifest, desktop/MIME integration and protected install/removal. Both math-profile packages passed staged checks, all four tools on synthetic inputs and native X11 welcome/preset captures; eight package guards pass. Vendor resources are excluded; build/install and limits — #11.
  • Software-profile core/file CI: 194 tests in debug and release per target on Linux x86-64, Windows x86-64 and macOS ARM, with Clippy clean. Cross-target checks found and fixed Windows CLI stack exhaustion and ARM NaN-sign differences; GUI/USB integration is outside this CI scope.
  • Raw Rust pickup circuit combination: 3,826 bit-exact comparisons with the native routine.
  • Pickup electronics: 2,322 raw responses + 387 primary fits + 282 secondary fits, 900 biquad checks, 280 complete tables and 252 smoothed-response comparisons. Synthetic fixtures pass debug/release; host integration is covered below.
  • Rust DSP packing/checksum, selected resonator conversions, 1,012 VX encodings and 315 scalar-setter cases match original-machine-code fixtures. 2,145 coefficient-writer cases, 1,320 complete pickup block cases and 1,008 simple-block cases pass. A further 600 biquad checks and 864 filter-block cases pass. All 17 DSP block paths now have parameter-to-word implementations: a further 2,304 EQ coefficient cases, 432 EQ block cases, 288 sitar cases, 432 body/string-tone cases, and 216 complete raw-body assemblies match the original instructions under the documented host-libm adapters. Scope is documented; the observed word map distinguishes 748 written words from 343 still-unknown words.
  • Pickup geometry and electric level primitives: 261 pole-offset outputs, 120 projection traces, 129 bundled-model placements, 72 synthetic host cases and 216 level/trim cases match the documented native paths — #2.
  • Offline Rust preset renderer: 240 Java mapping/overlay cases, 110 complete stateful native cases, 144 complete Java/native final-byte comparisons (including 24 special-combination cases), and 48 synthetic full-engine outputs pass. Extra regressions cover 300 fractional EQ mixes, 48 derived-frequency sitar cases, state retention, atomic errors and protected file output. External model resources remain private. Historical libm/audio parity remains pending; see host pipeline scope — #4.
  • Firmware inspection: both supplied images unpack with matching embedded MD5 and complete region coverage; 396 corrupt/truncated synthetic inputs rejected.
  • Offline Rust FORM/L6FF inspector (variax-firmware-inspect): bounded decompression, chunk and stream termination checks, region coverage, MD5 integrity and SHA-256 reporting. Synthetic rejection covers the same 390 prefixes and 6 damaged fields, plus wrong-product, wrong-version, oversized, overlapping, duplicate, and corrupted-image fixtures. MD5 agreement is not authenticity. No flashing or device API. Details: firmware containers — #17.
  • Offline artifact provenance: product, version, size, MD5, SHA-256, and a source URL only where one was captured. Variax 300/600 3.10 size and MD5 agree with Central release 1062. That agreement is not a signature. Details: artifact provenance — #12.
  • Linux Variax matching/routing investigation: interface ownership, the model-4 identity mismatch, an uninstalled patch, independently rebuilt module text and bounded Rust interpretation of compiled identity/startup/reset paths. Sends, timers and ALSA stay stubbed; scope and review — #21.
  • Simulated read-session fragmentation, cancellation, timeout, disconnect and cleanup failures; physical-access locks remain tested and enabled.

Pending implementation or validation

  • Historical macOS math parity and full Rust/native validation remain pending. The old uninitialized whole-generator/JNI baseline stays invalidated; use only the versioned candidates described in native runtime evidence.

  • Historical math-runtime parity, cross-target/default-rollout review of the experimental portable profile, and broader complete Rust/native validation — #5, #6. The synthetic-only core now passes Windows/macOS/Linux CI; other targets and full application integration remain pending.

  • Static firmware baseline map: Variax DSP56300-compatible load records, two boot-region streams and template/slot relationships; POD SHARC-compatible blocks/vectors, additive checksums and 486 named resource records. Exact ICs, opaque payloads, calibration and update/recovery behavior remain unverified. Ten synthetic layout guards; no firmware execution. Details: firmware map — #13.

  • Exact archived Variax 3.10 reference extraction: 50 compact templates, 60 full raw slots and 60 settings files; every program checksum validates and all 60 settings agree with the original Java decoder. Outputs remain private, save only to a new directory and are not calibration/current-unit backups. Details: factory extraction.

  • Saved Reso-5 metadata matches the firmware 3.10 default exactly: 6 models, 34 parameter type/value pairs, and identical settings-file bytes. Repeatable file-only comparison rejects other models/versions and preserves raw-bit distinctions. Earlier shipping state and the other stored presets remain unverified. Details: factory comparison — #15.

  • Additional calibration completeness/preservation, full-bank backup and physical restoration safety — #16, #23.

  • Bounded updater research oracle: 46 synthetic cases exercise six original host methods, two applicability helpers and three callbacks in private emulated memory. Code/preset/boot/POD encodings, checksums and selected timeout/error paths are corroborated with fake transports. No native DLL load, DLL entry point, device firmware execution or hardware access. This does not prove preservation or recovery safety. Details: update research.

  • Bounded static update/recovery map for the initial rig: product dispatch, transfer stages, integrity checks, native keep-preset flag path, documented POD safe-mode entry and explicit ROM/flash/calibration unknowns. This is research, not an updater or proven recovery path — #16.

  • Bounded in-memory Variax receiver stages and virtual POD/guitar metadata sessions, independent updater host-reference hashes, corruption/interruption tests and explicit unsupported recovery are implemented. See simulation scope — #18; physical recovery and backup readiness remain unverified under #22–#25. Rust container inspection is #17.

  • PODxt Live USB-component artifact and installed version remain unmet. A macOS Monkey userspace builder shape was recorded and was not sent. USB PID 0x4650 constructs an L6XTAudioControlMac and, when control slot +0x18 succeeds, virtually calls description slot +0x0c with that control. No symbol stub binds the setter. A later AcquireService(0) miss stores an L6KernelBrandiServiceMac, and the base acquire call returns that cache slot. Kernel GetVersion calls IOConnectCallScalarMethod selector 6. Sharedxt selector 6 builds the same five fields in a device request and forwards that pointer. That handler is not GetIOPVersion and does not identify a firmware image. The forwarded callee at slot +0x978 is not in the local archive, so the three returned bytes stay uninterpreted. That is a static branch. The five fields stay unapproved. The partial record is PODxt Live USB firmware. The installed USB version is unknown — #20.

  • Offline firmware-improvement feasibility assessment: specific host/firmware candidates, architecture and integrity constraints, calibration/preservation and recovery gaps. No deployable patch was generated; device changes are not ready. See assessment — #19.

  • Late-stage synthetic test expansion to at least 5x distinct samples, including body-dependent sustain/note decay, continuous string bends/vibrato, bends during decay and alternate/custom tuning in combination — #33. Audio behavior stays unverified without a suitable independent DSP/audio reference.

  • Readiness review, comprehensive backups and any future authorized hardware integration/write design — #22–#25.

  • Shared session, static registry, and capability matrix for one PODxt Live USB host with Variax 600 and PODxt Live roles. Guitar file open/save and per-family FORM/L6FF inspection are offline. PODxt Live tone open/rename is the offline librarian in #29. Physical enumeration, audio, MIDI, flash, reset, and driver actions stay locked. USB-component, installed-unit, and HD/JTV-69/POD HD500X versions stay unknown. Details: unified app — #26.

  • Standalone bridge control-record subset: source-qualified Windows x86 dispatch and Rust offline snapshot inspection; six owned synthetic methods pass standalone. No request builder, transport/session, attached-guitar identity or x64/macOS/physical compatibility is implemented. Complete-crate checks remain separately required; #61 stays partial.

  • #61 static package evidence, checked scopes v1–v5: 5 supplied originals, 10/10 inventoried copies matched the supplied sizes and SHA-256 references; 30 paired Windows/macOS Central JSON resources parsed to equal objects. Workbench HD 2.15's paired FactoryPresets.bun copies match by SHA-256; an offline bounded parser counted 12 banks × 5 presets, 12-byte PINF records, and eight typed parameter groups in each of 60 presets. Static macOS code confirms the WB2P/PINF/12-byte PARM contract and parameter index/type checks. A bounded Windows/macOS static comparison confirms the standalone LIST/TUNE serializer's six 8-byte STRG records: slots 0–5 map to ALTT indices 1–6 and float-typed values. Duplicate slots are last-wins; unknown children and malformed sizes fail; the macOS reader skips out-of-range slots while Windows fails. Six reproducible synthetic CPU-reference checks passed. Static UI wiring labels parameter indices 1–6 as String 1 (high E) through String 6 (Low E); this is not a physical routing observation. Controls declare −12..+12, while stored float units/range remain unproven. The standalone Windows x86 L6SM.sys PnP path also now has a static bounded removal trace: minor 0x02 conditionally detaches its saved lower device and deletes symbolic-link/device objects based on context flags. The matching standalone macOS driver is absent from the supplied set; attached-instrument identity, supported model list, interface/framing, complete disconnect cleanup and recovery remain unknown. The Variax USB driver INF identifies standalone bridge VID/PID 0x0E41:0x534D; Workbench HD's manual describes VDI editing and names HD Pro/HD 500, X3 Live/Pro, and Vetta II hosts. Central's Variax rows are update compatibility metadata, not editor support. Static Windows analysis finds Interposer.dll lifecycle/descriptor/capability/pipe/message exports; its PnP path parses USB VID/PID, admits VIDs 0x0E41 or 0x1235, then looks up PID in a manager tree. A traced parser reads caller-fed XML ProductId and library attributes and inserts base-16 product keys into that tree; the loader checks separate l6config and .xml string tokens, not a literal filename. The provider and Variax rows remain unresolved; supplied L6ConfigAmplifi.xml schema does not establish Variax support. Static .NET/C++ traces Central's ParamValueServer.initialize/start calls into native exports: initialize receives appVersionString, derives module-relative paths through GetModuleFileNameW, while start creates/obtains an engine context and dispatches through vtable 0x1047de8c to setter 0x1015a100, installing the createEngine callback bundle. Initialization also queries through the fourth createEngine callback. Static CIL maps it to EngineCLIWrapper.NativeStringParamValueReceiver; Central subscribes its SendString event to EngineInterface.setStringValue. The native argument values and any link to the Interposer XML parser remain unresolved. The updater builds and loads Interposer.dll from a path relative to its current directory. Its PE delay-import descriptor and name table bind nine named exports; the queue-out function-pointer slot resolves to InterposerMessage via the delay resolver and GetProcAddress. This is updater transport evidence, not a config provider, successful runtime exchange, or Variax editor proof. A raw LLVM x86_64 decode traces macOS product-ID branching to selection of Variax.json and a local helper. Static Objective-C metadata traces UpdaterV3Engine.getParamValueSetInterposerDelegate to the ParamValueServer singleton and its typed handler surface. Scanner teardown still has two unresolved vtable +0x40 dispatch targets and a selector/IMP mapping mismatch, detailed in the evidence report. Product selection, message framing, session/error behavior and editor parity remain unresolved. The tracked static provenance checker pins five private binary identities and qualified hashes for private code ranges without publishing vendor instruction bytes. All four configured command-v1 checks passed on the earlier head 7d42487a4a23332a4f5c33824a1f7ddf70c47008; that result applies only to that SHA and is not evidence for later heads. The PR status records checks for the current head; independent review and merge-result CI remain required. Synthetic coordinates and static source checks do not establish runtime callback, descriptor exchange, device compatibility or physical behavior. Vendor binaries were read statically but not executed. No device operation was performed; #61 remains partial. See cross-product package evidence.

  • All-device audio/recording compatibility remains unavailable. The schema-2 audio registry implements separate 224-row accounting and offline admission: one Linux PODxt Live static declaration, three denied maintenance labels and 220 explicit unresolved rows. Duplicate names/aliases remain separate; no live backend or stream is enabled. New Rust/CLI regression results require current-head checks on the associated PR. Parent VariaxWorkbench issue #96 remains open.

  • PODxt Live USB audio routing: offline Record Send 1-2 planning and synthetic S24_3LE buffer checks are in POD USB audio. GearBox modes are guide text. The PCM window is Linux v7.2.15 source text, with the same integers present in the installed snd-usb-pod module file. No stream was opened — #27.

  • Offline test recording in the same app: explicit synthetic or WAV input, Record Send route selection, integer peak meters, start/feed/stop/cancel, bounded chunks, and non-overwriting WAV plus sidecar saves. The source clock 78125/2 is kept and is not written as an integer WAV rate. Device capture stays locked. No DAW was opened. Explicit exact-clock 24-bit AIFF export preserves the rational and source PCM; eight files were independently decoded with SoX_ng. Other readers may round the rate; DAW imports are unverified. See AIFF scope — #69. Details: recording files — #28.

  • Offline PODxt Live tone librarian: FORM/L6PA, FORM/L6FS, and FORM/L6AS files with PINF target 0x0003000A, amp slot 0x00030000, and cab slot 0x00030001 on L6AS only. Asset names are the strings both archived GetAssetName copies return for this device. A 12-byte MINF shows the signed Java ordinal and the Read8() > 0 enabled flag. Reserved bytes stay raw. A 12-byte PARM shows its id, type, and raw bits. Unknown ids, other PARM lengths, and reserved bytes stay raw. No parameter unit or range was recovered, and values are not edited. Applied renames are tracked per tone, and quitting asks to save, discard, or keep editing before those edits are dropped. Nothing is sent. Details: POD presets — #29.

  • Bounded POD Farm 1.12 source recovery for #29: separately pinned original, embedded versions/architectures, seven selected component payloads with 832 regular files and seven Mach-O files; fresh extraction replay matched. Seven synthetic extraction-guard tests pass. Static PODxt-targeted packed-tone import and one typed scalar callback are connected; editor domains/units and the target-specific property provider remain unresolved. No vendor execution, new Rust editing behavior or hardware access; #29 stays incomplete.

  • Bounded POD Farm 2.58 inventory: digest-checked DMG, 16 packages, 62 Mach-O files and 1,120 structurally surveyed tone examples; clean extraction replay matched the private inventory. Ten synthetic file-reader guards pass. AU contains i386/x86_64; bundled VST/RTAS are i386. A native property trace identifies selector/ordinal bounds, state refresh and architecture-dependent copy layouts; 3,318 isolated two-leaf CPU cases pass on synthetic memory. No native vendor execution or device access. Domains and PODxt applicability remain unresolved. Evidence and partial scope — #40 remains open.

  • POD Farm 2.58 consolidated offline coverage: concrete host metadata/clamp and preset/state/conversion traces, eleven packaged resource IDs with conditional eligibility roles, VST parameter callbacks and capability/DAW-feasibility matrices. 69 owned synthetic test methods pass across eight suites (file audits and reconstructions, including six new packaged-header audits). Native parsing losses, partial mutation and unenforced spans are documented. Installed/authorized packs, final host automation bindings, scalar default/step/unit meanings, full-state/DSP equivalence and PODxt control applicability remain unverified. #40 stays partial and #29 value editing remains held. PR #84 is published; the required checks passed on the reviewed head. Changed heads require fresh check evidence; independent review and merge-result CI remain required.

  • Shared, device-specific firmware/backup/recovery/maintenance UI; real updates remain separately gated — #30.

  • Workbench HD/JTV-69 backends; other models added only with independent verification. An offline Rust inspector now bounds and reports generic FORM/WB2A structure, AINF, LIST/MDEF, MINF, and stored MDAT headers; its 8 synthetic parser/catalog tests pass. The CLI command is inspect-workbench-asset FILE; it does not decode payloads or claim model integrity/applicability. Static 2.15 tracing links ALTT to a compiled seven-entry typed table and index/type-only SetParam; reset seeds are zero, but stored units/ranges remain unknown. The JTV-family gate checks internal ID 0x70005 and the detected entry's L6Version, with a higher threshold when the MIDI port name contains POD HD. Catalog VariaxID keys are logical model-data lookup keys, distinct from asset model IDs and physical instrument identity; no JTV-69-specific variant link or device-specific asset filtering is recovered. No backend or JTV-specific editing is admitted. See loader evidence — #31.

  • Visual modernization remains a research proposal: the #41 visual study records static source evidence and three original SVG mockups, not implemented Rust behavior. The component follow-up adds 51-file static state/anchor counts, one font’s declared table metrics, two original annotated boards, layout/state contracts and five passing synthetic/static checks; rendered egui behavior and accessibility remain unverified. Exact reference framework version and runtime rendering/scaling remain unverified.

  • POD HD500X preset, model, effect, firmware-image, and USB-audio routing recovery. The Advanced Guide study now records historical documented processed-stereo USB routing/rate options and instrument-family constraints; Linux negotiation, formats and backend behavior remain unverified. Driver-text USB identity 0x0E41/0x4159 is recorded. Edit releases 9781 and 8559 answer HEAD with a login redirect, and HEAD of their release-id directories returns 404. No installer is local. A directory 404 can coexist with a public named object, and wider sources remain unknown. Two public 2013 manuals are hashed; their printed file suffixes disagree and are not a container. Independent H5EP and H5ES measurements are recorded, and their target bytes are not a registered product id. Details: POD HD500X — #32.

  • Unsaved offline recording takes are protected on quit alongside guitar/POD edits. Cancelled or failed saves retain audio; native-clock takes can be explicitly saved as exact-clock AIFF, kept or discarded. See recording files — #72.

Currently unavailable or not working

  • The image-default Windows x87 math substitution fails the Mac generator's strict startup-constant check; its complete-generator replay stops. Historical Mac parity remains unresolved, and the optional portable profile is experimental.
  • Guided check does not yet read switches/knobs, record audio or measure notes/bends/tuning. Future integration is #34; the current app rehearses prompts only.
  • Live device access is intentionally disabled, including identity/metadata reads. No runtime unlock exists.
  • Sending edited presets to the guitar, firmware flashing and factory restoration are unavailable; no live writer is implemented or authorized.
  • Live USB audio capture/playback is not implemented or validated; no audio stream has been opened. The offline test recorder does not open a device. See recording files.
  • POD Farm state interchange remains partial: a strict Rust XFP6 envelope inspector/save-copy preserves opaque prefix/section/gap/tail bytes, with six synthetic Rust test methods requiring exact-head CI evidence. Nested conversion, pack/automation applicability, host binding/application and DSP equivalence remain unverified; this is not a functioning plugin adapter or PODxt value editor.
  • PODxt Live parameter units, ranges, and value editing are not recovered. The archived Workbench jars name a Sharc editor that is not present, and the only L6ParamDefinition ranges belong to Variax. The POD Farm inventory supplies static leads, not applicable PODxt domains, pack entitlements, lossless conversion or plugin adaptation. Firmware region-17 editing and sending a tone to a device are not implemented. See POD presets and POD Farm limits — #29, #40.
  • POD HD500X preset editing, firmware images, and live audio are not available. No preset container, model id, effect id, or firmware image was recovered. Printed manual suffixes are not a byte layout. Independent file bounds do not identify the product. The INF 48000 Hz / 24-bit values are driver defaults with EngineFormatLock 0. See POD HD500X — #32.
  • Full PODxt Live/Variax firmware and audio/DSP emulation is not implemented. Existing host-library emulation is not device emulation.
  • The earlier experimental full-DSP read timed out; no complete DSP/bank/calibration backup was obtained. The captured “Reso-5” file is one preset's settings only.
  • Earlier Linux observations had no Variax ALSA port. The model-byte mismatch is diagnosed and an uninstalled patch is verified offline; live connectivity remains unverified.
  • No complete authenticated Variax 600 factory bank or proven recovery path is available.
  • The original Java reader rejects the tested odd-length unknown bank extensions with EOFException; Rust preserves them, but Java compatibility is unverified for such files. Entirely empty bank/collection GUI editing and bank/preset creation/deletion are unavailable. See bank limits.
  • Native Wayland screenshot validation timed out inside Mesa buffer swap in this session; the Linux package launcher uses the tested X11 path. Wayland rendering and native Windows/macOS GUI checks remain unverified; see package validation.
  • Original-library reference tests and illustrated artwork need separately supplied private artifacts. The Linux package opens without them; Windows/macOS desktop packages are not implemented.

Run the offline editor

cargo run --features gui --bin variax-workbench-gui -- --display x11

Open a file with Open or --preset FILE. Optional extracted artwork can be selected with Artwork folder or --artwork EDITOR_IMAGE_FOLDER; see package paths. The editor supports body/pickup selection and placement, tuning and electronics controls, undo/redo, and saving a new copy. Originals and existing output files are protected. A saved preset cannot yet be sent to the guitar. Offline host generation is implemented; historical-runtime, broader full-engine and audio validation remain pending.

The build's coefficient math profile can be inspected with math-profile. The optional software profile and its offline checks are documented.

Offline command line

cargo run --bin variax-workbench -- inspect INPUT.vwm
cargo run --bin variax-workbench -- rename INPUT.vwm COPY.vwm "My Guitar"
cargo run --bin variax-workbench -- set INPUT.vwm EDITED.vwm 02010001 200000 float 2.5
cargo run --bin variax-workbench -- inspect-bundle BANK.vwb
cargo run --bin variax-workbench -- extract-preset BANK.vwb EXTRACTED.vwm 0 0

Output paths must not already exist. set takes hexadecimal model-local IDs; explicit edits use the recovered parameter domains and body restrictions. Imported unknown and dormant values remain lossless. decode FIRST.syx SECOND.syx decodes two captured metadata replies. preview prints recovered request bytes without transmitting them.

Verify

cargo test --all-features
cargo clippy --all-features --all-targets -- -D warnings
cargo run --bin variax-firmware-inspect -- --self-test
python scripts/verify_preset_compatibility.py
python scripts/verify_bundle_compatibility.py
.tools/python/bin/python scripts/verify_native_pack.py
.tools/python/bin/python scripts/verify_native_resonators.py
.tools/python/bin/python scripts/verify_native_pickups.py
.tools/python/bin/python scripts/verify_native_bodies.py
.tools/python/bin/python scripts/verify_body_rust.py
.tools/python/bin/python scripts/verify_body_plans.py
.tools/python/bin/python scripts/verify_body_interpolation.py
.tools/python/bin/python scripts/verify_pickup_combinations.py
.tools/python/bin/python scripts/verify_pickup_response.py

Reference checks require the locally extracted original application. REA was built in ../rea; scripts/rea selects the local Node/Java/Ghidra toolchain. Original installers, proprietary extracts, artwork, captures and portable tools remain local and ignored. No updater or installer was executed.

Details: preset formats, body definitions, artwork, factory/firmware investigation, initial investigation, native analysis.

Pending work is tracked in GitHub issues and the dependency-linked backlog. The default branch is master.

The VariaxWorkbench project tracks execution order, phase and status, with native issue dependencies. Work proceeds one issue at a time; hardware access stays locked.

About

Rust offline editor and reverse-engineering toolkit for the original Variax 600 through PODxt Live; hardware access locked.

Resources

Stars

2 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages