gcc: record verified upstream sources and the port sequence

Latest upstream RELEASE tag is releases/gcc-16.1.0; '16.1.1' seen on distros
is a packaging/snapshot string, not an upstream release. Both
gcc-16.1.0.tar.xz and gcc-15.3.0.tar.xz confirmed reachable on ftp.gnu.org
(HTTP 206 on a range request). The Redox gcc fork carries no upstream
release tags, so the pristine base must come from ftp.gnu.org or
gcc-mirror.

Records the ordered next steps and marks them explicitly NOT STARTED, so the
extraction is not mistaken for a working port. The dominant cost is step 5 --
the full C++ tree rebuild forced by the libstdc++ ABI change -- not the
13-file port itself.

local/patches/gcc-redox-port/README.md

@@ -79,3 +79,37 @@ Inference, not a maintainer statement: Redox's own toolchain is
 Clang/LLVM-21-centric and GCC is the secondary C/C++ path, so upgrade pressure
 is low. They have jumped versions before (8.2.0 → 13.2.0), which argues the gap
 is demand-driven rather than a technical blocker.
+
+## Verified upstream sources (checked, not assumed)
+
+Latest GCC **release tag** is `releases/gcc-16.1.0`. Note "16.1.1" seen on
+distros (e.g. Arch) is a packaging/snapshot string, not an upstream release.
+
+Tarballs confirmed reachable (HTTP 206 on a range request):
+
+    https://ftp.gnu.org/gnu/gcc/gcc-16.1.0/gcc-16.1.0.tar.xz
+    https://ftp.gnu.org/gnu/gcc/gcc-15.3.0/gcc-15.3.0.tar.xz
+
+Prefer the tarball over `git clone` of gcc.gnu.org (~2 GB) — it suits the
+offline-first build model and matches how other external recipes vendor sources
+(`tar =` + `blake3 =`).
+
+`gitlab.redox-os.org/redox-os/gcc` carries **no** upstream release tags, so the
+pristine base must come from ftp.gnu.org or gcc-mirror, not from the Redox fork.
+
+## Next steps (in order)
+
+1. Vendor `gcc-16.1.0.tar.xz`, record its blake3, create
+   `local/recipes/dev/gcc16/recipe.toml` (model on `recipes/dev/gcc13`).
+2. Apply `files/` + the arms in `registration-hunks.txt`; regenerate `configure`
+   from `crossconfig.m4` with autoconf.
+3. Parameterize the hardcoded `13.2.0` in `mk/prefix.mk` (at minimum the
+   `lib/gcc/$(GNU_TARGET)/13.2.0/include/limits.h` path).
+4. Build the cross compiler; stage into `prefix/`. Expect relibc header
+   fallout (see risks above) before anything else compiles.
+5. Full C++ tree rebuild for the libstdc++ ABI change — Qt6, all KF6, Plasma,
+   Mesa C++. Budget this as the dominant cost, not the port itself.
+6. Only then retry `kwin`, whose 16 `std::ranges::to` files are the reason for
+   the whole exercise.
+
+**Not started.** Steps 1–6 are unvalidated; nothing below step 0 has been run.