Skip to content

Release/0.5.2 dry run - #1

Merged
thomchantra merged 116 commits into
mainfrom
release/0.5.2
Nov 13, 2026
Merged

thomchantra merged 116 commits into
mainfrom
release/0.5.2

Conversation

@thomchantra

Copy link
Copy Markdown
Owner

No description provided.

thomchantra and others added 30 commits November 10, 2026 01:50
…lies it)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…, bit-exact audio + telemetry)

Scenarios cover every shaper x oversampling mode, compressor/EQ/sidechain/bypass/delta/match variants, ragged blocks, the
tick API, automation, other sample rates. The fixture was recorded from the v0.5.1 tag, not from this tree. The example
renders to disk and diffs two render folders for the baseline-vs-refactor comparison.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… s match accumulator)

New modules only; the engine is moved onto them in the next commit. Not wired into lib.rs yet.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…qSection); Engine now only composes them

Pure refactor: all 101 golden scenarios are bit-identical to v0.5.1 (audio and telemetry), all unit tests pass.
- saturator.rs: HP/LP split, shaper + oversampling, DC block, tone, mix, SAT OUT; owns its delay rings and level
- compressor.rs: detector, key mix, gain computer, ballistics, auto release, GR distortion (GainComputer moved here)
- keyinput.rs: external key gain > HP > LP > latency alignment
- eq.rs: EqBank is now one 5-filter stage; EqSection adds the smoothed enable crossfade and output gain
- SatParams / CompParams are per-block parameter slices (Params::sat(), Params::comp()); Params stays the flat
  SatComp-level set so the plug-in and UI are unchanged
- Engine keeps its public fields and methods (meters, match, scope) as a facade; block meters are mirrored per tick

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…mp())

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ames); editor closures become init_ctx / run_frame

The test runs the real frame function in an egui context against a fake host, drives it with a scripted mouse (sliders,
EQ page, About/Appearance/Changelog, menus, pencil view, scope, presets, a hi-dpi pass) and hashes every frame's draw
commands plus the host parameter writes. Audio-side meters and the spectrum tap are fed deterministic signals.
Limits: nih-plug keeps its parameter setters crate-private, so values are not echoed back (parameters stay at their
defaults); scope ring data is not fed (private writer).

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…s, init_ctx and FFT (moved out of the plug-in)

First step of the UI split so SatXComp / SatEQ can share the drawing layer. Files moved with git mv (canvas.rs, fonts);
gen_palettes.py now writes satcomp-ui/src/canvas.rs. UI golden (460 frames) unchanged.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…trait: cell, cell2, cell_o, btn, sqind, sq_tog, drag/wheel/double-click, slot addressing)

The plug-in implements Widgets for its per-frame Gui (slot -> ParamPtr, value formatting, host context, edit mask);
State derefs to Kit so page code still reads self.mx / self.press unchanged. SatXComp and SatEQ implement the same trait with
their own parameter tables. UI golden (460 frames) unchanged. Not yet moved: menus, About/Appearance/theme editor, presets.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… not invalidate the UI golden

The fixture was re-recorded from v0.5.1 carrying the same one-line test stub, and the refactored tree matches it on all
460 frames (the only difference from the previous fixture is the header version text).

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…y, README, release notes

The UI golden fixture is updated for the new changelog text: only the 17 Changelog-page frames differ from the v0.5.1
recording, every other frame is unchanged.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…two regression nets and how to diff against a release; AU shelved; sibling plug-ins live in this repo

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ld against the previous release in the host); documented

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…shed release is v0.5.1)

Crate versions, in-plugin changelog entry, README, FOUNDATION and the release notes file (docs/releases/v0.5.2.md) renamed.
UI golden re-recorded: only the 10 Changelog-page frames moved (the entry's version label).

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…e generic in filter count (SatComp keeps 5)

For SatXComp's 5-band EQ (+2 tilt filters) and SatEQ. SatComp audio golden unchanged (bit-identical).

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… 0.05 dB, -6 dB at each split, glides while dragged)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…mix/gain, solo/delta/on, 5-band EQ + tilt, mix, out gain, gain match, equal-loudness bypass, spectrum taps)

Tests: neutral level, band isolation, solo / off passthrough, EQ boost + latency report, all oversampling modes.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ools/gen_xparams.py, plug-in shell around XEngine, ids thomchantra

VST3 class id thomchantraSatXC, CLAP id com.thomchantra.satxcomp, vendor thomchantra. Parameter ids use lo_ / mid_ / hi_ and eq1..eq5
prefixes; UI slot numbers: globals 1-8, bands 10+24b, EQ 90+5i, EQ globals 115-119. Tests: generated defaults equal the engine defaults,
every slot resolves and is unique, ids unique. pluginval strictness 5 passes. Hosts show their generic UI until the editor lands.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…in updates; CI builds each plug-in; SatXComp starts at 0.1.0

build.yml: a plan job maps the tag to the plug-ins (v* = every plug-in in the workspace, satxcomp-v* / sateq-v* = just that one);
bundle matrix = plug-in x (macOS universal, Windows x64); zips are named with each plug-in's own Cargo version; the release job runs for
any tag. release-notes.yml applies docs/releases/*.md for every tag form. FOUNDATION documents the scheme; v0.5.2 notes get the bundle table.
SatEQ is listed in the notes but has no code yet: the plan job only includes plug-ins that exist.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…locked)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…B, log-spaced tilt-compensated columns, log frequency axis helpers)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ow and the EQ page

Main page (logical 560x384 grid, window default 840x576): header with icon (click = next scheme, right-click = light/dark) and
out meter; x1/x2/x4 oversampling (saturators only), EQ page toggle with module on/off square, DELTA, BYPASS; spectrum viewport
(output in the UI colour, what the compressors removed in the accent, output above dry in the inverted accent) with two draggable
dotted crossover lines and L/M/H band markers showing each band's GR; X-OVER 1/2, L.AHEAD, PK/RMS; band row (LOW/MID/HI tabs with
on squares, DELTA, SOLO of the active band); per-band saturator (shape selector, DRIVE, transfer plot, BIAS/TONE/SAT MIX/SAT GAIN),
GR meter and compressor cells in the draft's order; output column (gain match, IN>OUT, MIX, OUT GAIN).
EQ page: spectrum + summed response curve with five draggable numbered nodes (wheel = Q), band toggle row (type name or "EQ n"),
per-band type glyph grid (all 8 filter kinds via the per-band lists) with Q / FREQ / GAIN, PIVOT / TILT / EQ MIX / EQ GAIN.
Built on satcomp-ui's Kit and Widgets. Headless smoke test drives both pages against a recording fake host.
Not yet: menus, About/Appearance/theme editor, preset bar, custom-curve draw view, sidechain, resize grip.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… updated the saturator's SAT OUT after the first snap); regression test

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…hey do: the no-ops were in the editor)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… EQ mix, EQ gain never moved) and node/crossover drag ids colliding with real slots; regression test

kit.act < 100 meant "slot drag", but slots run to 119, and viewport drags used 101..107 which are real slots (moving EQ2 drove EQ3 gain, EQ3 lit a Q cell). Viewport drags now start at 1000.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…kwitz-Riley), flat-sum tests for every slope pair

The 12 dB/oct split only sums flat with its high side inverted; the engine sums bands with Crossover3::signs() so the bands keep natural polarity for the saturators.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… dB/oct), 103 parameters

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…t, value right, bar on the bottom edge

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…lls, right column 140 wide

DELTA / SOLO of the active band sit next to the band tabs, L.AHEAD is a small row at the end of that row, drive is a small row above a taller transfer plot, X-SLOPE 1/2 under X-OVER 1/2, EQ page columns widened. Layout comes from named constants instead of literals; smoke tests moved to the new coordinates and cover the new controls. Default window 1020 x 720.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… for an EQ IN>OUT readout; published through Shared

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…pe button groups, band tabs flush with DELTA/SOLO, sat module on the grid, EQ polish

Viewport back to 128 high, PK/RMS directly above the crossover groups, L.AHEAD on both pages. Shaper selector + drive row flush (one cell row), plot on rows 2-4. Band and EQ squares show an X when bypassed (EQ bypass dims the EQ page, an off band dims its cells). EQ: stroke lines between the type buttons, double-click resets a band (toggle or node), EQ IN>OUT row, dashed pivot line with pivot drag and sideways tilt drag, TILT readout.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
thomchantra and others added 29 commits November 13, 2026 13:59
…ach strip measures its band's loudness change across the saturator (input vs the result before SAT GAIN, after the band confine), the engine runs a 3 s matcher per band (request_sat_match / take_sat_result, a second click cancels), the plot's bottom right corner shows SAT +x dB (MEAS n s while measuring) and clicking it measures the selected band and writes its SAT GAIN; shared fields per band; DSP test (drive 70 on a low tone) and a UI test

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…, DRUMS_UTubes, Deep n Bright, Mildly Cooked with its custom curves; values rounded; limitertest1 and the stale pre-4-band Sub Sidechain left out); test for names, ids and ranges

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ns on a fixed grid in absolute bucket time, only complete ones drawn, read position eased toward the head and re-based on editor reopen; scope_col is pure and unit-tested), replacing the per-frame regrouping that flickered with host-block chunks

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…d VU and the GR meter so the editor's 4 x 3 grid no longer overlaps them

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…upward mode (CompParams up / up_range, default off and bit-identical: the SatComp audio golden is untouched): the gain computer mirrored around the threshold lifts what is below it with the same ratio and knee, capped at RANGE, the lift rises with release ballistics and is pulled back with attack so loud transients are never boosted; CompOut.lift. SatXComp: params <band>_up (slot band+23) and <band>_up_range (slots 160-163), generated by tools/gen_xparams.py (the generator now also emits from_defaults, so regeneration reproduces params.rs); engine/strip report the lift (gl_pk). UI: the second compressor column is now COMP DOWN / UP over RANGE, KNEE over K.SHAPE (stacked pairs), then GR DIST and SPEED; in UP mode the GR meter is centre-tapped, lift grows upward in the spectrum's gain colour (inverse of the accent), reduction downward in the accent, the number window shows the lift. Tests: compressor (down unchanged, lift value and cap, quick give-way), engine, UI

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ain compressor), UPWARD (lift below the threshold) and UP/DWN (both, lifting below and reducing above the same threshold). CompParams gets `down` (default true, so the SatComp audio golden stays bit-identical: lift 0 leaves the old gain expression exact); the band param <band>_up is now an int 0..2 (0 / 1 keep their meaning from the bool); the switch cycles on click and steps with the wheel; in UP/DWN the centre-tapped meter gets a second number window at the bottom for the reduction; tests for the cycle, the both mode and the unchanged upward / downward cases

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…chevron: SET AS DEFAULT saves this plug-in's own default (defaults-<plugin>.json), its chevron opens a one-item menu SET AS GLOBAL DEFAULT (the old shared defaults.json); a new instance reads its own default, else the global one, else the factory look (load_defaults_for); RESET works the same way (main part clears this plug-in's default and takes what a new instance would now get, its chevron item clears the global one; the label reads RESET TO GLOBAL while a global default exists, cached so no disk read per frame); the menu takes the click before the page under it; font pool 11 with Inconsolata and Courier Prime (Regular, tuned sizes 1.14 / 1.15); SatLim test walks the whole flow; SatComp golden regenerated

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…nd seven schemes are appended (Brick, Bob, Carbon, Olive, Jade, Honey, Stealth; 27 built-ins, values from the owner's themes.json; docs/colorscheme.md -> tools/gen_palettes.py as before); a saved user theme now wins over the stored index when a project is restored (the built-in table grew past the old user-theme indices, so index 20+ must not be read as a built-in), a user theme whose name is now a built-in resolves to the built-in, and user themes shadowed by a built-in name are left out of the list (themes.json is not touched until the next save); dice button next to COLOR SCHEME rolls a random scheme (never the current one) and scrolls the list to it; tests for the migration, the new schemes and the dice; SatComp golden regenerated

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…tead of an exact scroll offset (the page clamps the scroll to the last full screen)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…: the logo icon inverts under the pointer (UI-coloured plate with a BG glyph, or back to the outline in light mode) and the app title gets a UI-coloured highlight rectangle behind it with the text in the background colour (all three plug-ins); SatComp golden regenerated; SatLim test

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… through (the delay line already held ~2 ms the detector never saw and the gain state restarted at unity: 75 samples at the raw level, +10 dB on a hot tone; the gain now starts from what keeps the loudest of those samples at the ceiling and no target may rise above it until they have left the line); optional doubled detector grid (set_hi / TruePeak::with_precision: 8x at 48 kHz-class rates, 4x at 96, 2x above), off by default so SatComp and SatXComp are bit-identical (audio golden unchanged); tests for both

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
…ter is clipped / folded, all inter-sample peaks: on the amen loops the transparent case holds the ceiling within 0.04 dB on a 32x reference meter, 0.07-0.10 at 4x); the CLIP reduction readout is how far the pre-clip peak stood above the ceiling (the plain pre / post ratio read 30-90 dB with oversampling on, where the two differ by the filters alone; same number as before at x1); delay rings 2048 so bypass and delta stay aligned at 352.8 / 384 kHz; tests

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
…ashlogs, docs/SatComp_spec.md now under docs/internal) and ignore the owner's working folders (docs/internal, docs/crashlogs, docs/feedbacknotes, wavtest, screencaps)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
…ext)

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
….PK with L.AHD off too: the zero-latency mode only holds the sample peaks, and the readout showed -1.0 while the true peak of a folded amen loop was +5 dB; test; SATLIM doc

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
…scaling (dry input scaled to the processed RMS, up to +10 dB peaks with a hot saturator and no limiter) and the BYP readout are removed; bypass returns the latency-aligned dry input at its own level, like SatLim; test now checks a sample-exact delayed copy with drive 80; DSP golden re-recorded (audio changes only in the bypass scenario, meter hash drops bg_gain); docs and v0.5.2 notes

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…(lift, in the gain colour); gain colour keeps the inverted hue on dark/saturated themes (Barracuda, Cadmium, Azure, Uranium, Midnight no longer fall back to white): grey schemes detected by hue (grey accent or all swatches within 30 degrees), otherwise the inverse is lightened toward white/black until it stands off the background; shared fn in satcomp-ui, test over all palettes

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…aturated (Barracuda, Uranium, Apricot and other single-hue themes keep their inverted hue); only Ashen-like schemes use white/black

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… (docs/abouttextcopy.md, now committed with its edits): SatXComp gets the viewport and band-handle legend, all three the shared controls (GR meter range, clip meter reset, theme and dark/light icon); changelog entries for the frozen features (SatXComp v0.1 with upward modes and the lift bar, SatComp v0.5.2 plain bypass, SatLim v0.1 8x true peak); SatComp UI golden re-recorded, only the about / appearance / changelog frames moved

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…ts of all three (standalone with a mock input, nothing of the hack committed; SatLim fed one of the amen loops); docs/INVENTORY.md tracks versions, ids, milestones and the release checklist; v0.5.2 notes describe the two new plug-ins (4-band SatXComp, SatLim; the SatEQ rows are gone, it only exists as a pitch); OFL licences for the five fonts added in 0.5.2

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…lders (VST3/<plugin>.vst3, CLAP/<plugin>.clap, INSTALL.txt per OS, LICENSE, the manual) so the Windows .vst3 folder is no longer a puzzle; the manual PDF for the plug-in's version is attached to the release when docs/manual/<Name>-Manual-v<ver>.pdf exists; docs/manual: SatComp / SatXComp / SatLim markdown sources, build.py (WeasyPrint: A4, black pages, white Fira Code) and the first PDFs; README and INVENTORY describe the layout

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…bout tab (each plug-in embeds docs/manual/<Name>-Manual-v<version>.pdf, so a version bump without a rebuilt PDF fails to compile; the button writes it to the temp folder and opens it with the default viewer, online folder as fallback); SatXComp flow shows the band's compressor mode (SAT>COMP / UP / UP/DWN) and lights on lift as well as reduction; flow blocks are switches (SatXComp: bands, EQ, LIM, SIDECHAIN; SatComp: PRE EQ, POST EQ, LIM) with an accent hover; test; SatComp UI golden re-recorded (about / appearance / changelog frames only); manuals: support links (GitHub, Ko-fi), SatLim screenshot fits page 1, About notes

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…gnal flow and the theme gain colour; Ko-fi sponsor button (.github/FUNDING.yml)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
… CLAP/, INSTALL.txt from docs/install/INSTALL-Linux.txt)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…t) in all three plug-ins). The UI golden stepped through the machine's own user presets, so it only matched on the machine that recorded it: CI (clean home) differed in the 12 preset / logo frames. SatComp UI golden re-recorded (those 12 frames), verified under an empty HOME

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…and Ko-fi (all three systems); Linux zip gets install.sh (user or --system install, --uninstall) and its INSTALL.txt explains the hidden ~/.vst3 and ~/.clap folders and Ctrl+H

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…wap interval defaults to off on Linux (a blocking vsync swap on top of baseview's 15 ms frame timer is the suspect for the stuttering meters and the drunken transient scope on X11), on elsewhere; SATCOMP_VSYNC=1 / 0 forces it for A/B; SATCOMP_FPS=1 prints frames per second, average and worst frame time to stderr once a second

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
…, INVENTORY) with the meter stutter as a known issue; the GL vsync default is back to on everywhere (turning it off on Linux did not help; SATCOMP_VSYNC / SATCOMP_FPS stay for diagnosis)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@thomchantra
thomchantra merged commit 0767d5f into main Nov 13, 2026
12 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant