@gjsify/gtk-runtime-darwin-arm64 0.51.1 → 0.53.0

package/README.md

@@ -5,6 +5,10 @@ Batteries-included, **relocated** GTK / GObject-Introspection runtime bundle for
 (GLib · GObject · Gio · cairo · Pango · Graphene · Gdk) with **no Homebrew GTK**
 installed on the host — Phase 2 of cross-platform node-gi.
 
+Every image in the bundle loads on **macOS 15 or newer** — the floor all gjsify darwin
+binaries share ([ADR 0074](../../../docs/adr/0074-one-declared-macos-floor.md)); the
+builder fails on a Homebrew bottle poured for a newer OS.
+
 Platform-gated (`os: ["darwin"]`, `cpu: ["arm64"]`), tier 3 (experimental). On any
 other platform npm skips it and `@gjsify/node-gi` falls back to a system/Homebrew
 GTK. The heavy `gtk/` payload is **not committed** — it is built on a macOS CI
@@ -192,7 +196,7 @@ Still not collected: the `gdk-pixbuf` loader modules + `loaders.cache` (they are
 dylibs needing `@loader_path` relocation from a nested dir, unlike win32's flat DLL
 copy), so SVG symbolic icons can render blank; and Fontconfig config/cache, which
 macOS text rendering does not need (Pango uses the CoreText backend). Both are
-tracked in `status/open-todos.md`.
+tracked in `status/open-todos/README.md`.
 
 **Env-free mechanism (macOS):** because dyld captures
 `DYLD_FALLBACK_LIBRARY_PATH` only at launch, node-gi re-execs once with it set (see

package/gtk/THIRD-PARTY-NOTICES.md

@@ -128,6 +128,8 @@ each other from inside the bundle instead of from the build machine.
 - `libgstbase-1.0.0.dylib` — gstreamer 1.28.6_3
 - `libgstcoreelements.dylib` — gstreamer 1.28.6_3
 - `libgstflac.dylib` — gstreamer 1.28.6_3
+- `libgsticydemux.dylib` — gstreamer 1.28.6_3
+- `libgstid3demux.dylib` — gstreamer 1.28.6_3
 - `libgstisomp4.dylib` — gstreamer 1.28.6_3
 - `libgstmpg123.dylib` — gstreamer 1.28.6_3
 - `libgstmulaw.dylib` — gstreamer 1.28.6_3

package/gtk/manifest.json

@@ -5,11 +5,11 @@
   "generatedFrom": "/opt/homebrew",
   "dylibs": 81,
   "typelibs": 48,
-  "libBytes": 65584766,
+  "libBytes": 66380286,
   "typelibBytes": 3085900,
   "dataBytes": 28135370,
   "licenseBytes": 1173469,
-  "totalBytes": 97979505,
+  "totalBytes": 98775025,
   "dylibList": [
     "libFLAC.14.dylib",
     "libX11.6.dylib",
@@ -277,6 +277,7 @@
       "GObject",
       "Gio",
       "GModule",
+      "GIRepository",
       "cairo",
       "Pango",
       "PangoCairo",
@@ -306,7 +307,7 @@
     "attribution": "per-binary",
     "components": 56,
     "texts": 111,
-    "binariesCovered": 119,
+    "binariesCovered": 121,
     "binariesModified": true,
     "modifications": [
       "`install_name_tool -id` / `-change`: every install name and every reference to another library in this bundle rewritten to `@loader_path/<leaf>`, with the prefix climbing back to `lib/` for the modules that sit below it — `../` for the GStreamer plugins, `../../` for the GIO modules, `../../../` for the gdk-pixbuf image loaders (references to /usr/lib and /System are untouched).",
@@ -315,7 +316,7 @@
   },
   "windowingData": {
     "pixbufLoaders": 13,
-    "gstPlugins": 24,
+    "gstPlugins": 26,
     "gioModules": 1,
     "schemas": true,
     "iconThemes": [

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@gjsify/gtk-runtime-darwin-arm64",
-  "version": "0.51.1",
+  "version": "0.53.0",
   "description": "Batteries-included, relocated GTK/GObject-Introspection runtime bundle for macOS arm64 — lets @gjsify/node-gi load gi:// namespaces (GLib/GObject/Gio/cairo/Pango/Graphene/Gdk) with NO Homebrew GTK. Platform-gated (darwin/arm64 only).",
   "type": "module",
   "license": "SEE LICENSE IN gtk/THIRD-PARTY-NOTICES.md",
@@ -74,8 +74,12 @@
       ],
       "gaps": [
         {
-          "format": "AAC (in M4A)",
-          "why": "`isomp4` demuxes the container and `audioparsers` supplies `aacparse`, after which the stream reaches no decoder at all. The two elements that would decode it are `faad` (gst-plugins-bad, GPL) and `avdec_aac` (libav, whose whole closure these bundles refuse), so shipping one is a redistribution decision for whoever ships the product rather than this bundle's."
+          "format": "AAC (M4A)",
+          "why": "`isomp4` demuxes the container and hands the decoder raw AAC directly, after which nothing decodes it. The two elements that would are `faad` (gst-plugins-bad, GPL) and `avdec_aac` (libav, whose whole closure these bundles refuse), so shipping one is a redistribution decision for whoever ships the product rather than this bundle's. win32 claims this format instead, through `mfaacdec` — the OS's own decoder, wrapped by the same `mediafoundation` plugin that closed MP3 there (ADR 0056 § 7); that route does not exist here, because there is no Media Foundation on macOS."
+        },
+        {
+          "format": "AAC (ADTS)",
+          "why": "`audioparsers` supplies `aacparse` for a raw ADTS stream, after which nothing decodes it — the same `faad`/`avdec_aac` exclusion as the `AAC (M4A)` gap. win32 does not claim this shape either, even though it has `mediafoundation`: measured on CI, `mfaacdec` decodes the M4A shape but stalls (no sample, no EOS) on a bare ADTS stream through `decodebin3` (ADR 0056 § 7). So this format is a gap on every target today."
         }
       ]
     }