fix: automatic patch for top-k.cu to file ggml build error

2 files changed, 113 insertions, 6 deletions

diff --git a/app/backends/audiocpp.py b/app/backends/audiocpp.py
index 7a8eb75..9cb8e24 100755
--- a/app/backends/audiocpp.py
+++ b/app/backends/audiocpp.py
@@ -95,6 +95,12 @@ TASK_VDES = "vdes"
 AUDIOCPP_DIR_NAME = "audio.cpp"
 AUDIOCPP_GIT_URL = "https://github.com/0xShug0/audio.cpp"
 
+# ggml build patches shipped in this repo and applied to the (gitignored)
+# audio.cpp checkout before building, so a fresh clone survives known ggml
+# build bugs the audio.cpp fork has not re-vendored yet. See
+# apply_ggml_patches() below.
+PATCH_DIR = Path(__file__).resolve().parent / "patches"
+
 # Sentinel returned by tui.confirm (via its cancel_value) when the user
 # presses Esc on an overwrite prompt to go back to the wav-directory browser
 # instead of aborting the wizard.
@@ -1358,18 +1364,24 @@ def _wizard(stdscr, args: argparse.Namespace, parser: argparse.ArgumentParser
         audiocpp_dir = find_local_checkout()
     if audiocpp_dir is None:
         target = APP_DIR / AUDIOCPP_DIR_NAME
-        rc = taskview.run_steps(stdscr, "Clone audio.cpp", [taskview.TaskStep(
-            f"Cloning audio.cpp into {target}",
-            lambda emit, cancel: common.git_clone(
-                AUDIOCPP_GIT_URL, target, emit=emit, cancel=cancel))])
+        rc = taskview.run_steps(stdscr, "Clone audio.cpp", [
+            taskview.TaskStep(
+                f"Cloning audio.cpp into {target}",
+                lambda emit, cancel: common.git_clone(
+                    AUDIOCPP_GIT_URL, target, emit=emit, cancel=cancel)),
+            taskview.TaskStep(
+                "Apply ggml build patches",
+                lambda emit, cancel: apply_ggml_patches(
+                    target, emit=emit, cancel=cancel)),
+        ])
         if rc == 130:
             # Cancelled from the task view: abort the wizard quietly.
             return None
         if rc != 0:
             raise _TuiError(
-                f"git clone failed (exit {rc}). Clone "
+                f"audio.cpp setup step failed (exit {rc}). Clone "
                 f"audio.cpp manually: git clone "
-                f"{AUDIOCPP_GIT_URL} {target}")
+                f"{AUDIOCPP_GIT_URL} {target}, then re-run")
         audiocpp_dir = target
     resolve_checkout(audiocpp_dir)
     first = _after_families()
@@ -1818,6 +1830,75 @@ def find_build_script(audiocpp_dir: Path) -> Optional[Path]:
     return candidates[0] if candidates else None
 
 
+# Each entry pairs a shipped patch with the vendored file it touches and a
+# regex marker proving the fix is already present (so the patch is skipped
+# idempotently once applied, or once the fork re-vendors a fixed ggml).
+GGML_PATCHES = [
+    {
+        "file": "ggml-top-k-cuda-iterator.patch",
+        "target": "external/ggml/src/ggml-cuda/top-k.cu",
+        "marker": r"#\s*include\s*<cuda/iterator>",
+        "label": "top-k.cu: add #include <cuda/iterator> (CCCL 3.x build fix)",
+    },
+]
+
+
+def apply_ggml_patches(audiocpp_dir: Path, *, emit=None, cancel=None) -> int:
+    """Apply the shipped ggml build patches to an audio.cpp checkout.
+
+    Idempotent: a patch whose marker already matches its target is skipped
+    (it is either already applied, or the fork re-vendored a fixed ggml). A
+    patch that no longer applies because the vendored file changed shape is a
+    loud, non-interactive failure — the build is aborted so the user
+    re-evaluates the patch instead of hitting a known nvcc break minutes
+    later. Returns 0 when every patch is applied or already present, 1 on
+    drift, 130 when cancelled.
+    """
+    for patch in GGML_PATCHES:
+        if cancel is not None and cancel.is_set():
+            return 130
+        target = audiocpp_dir / patch["target"]
+        if not target.is_file():
+            print(f"[INFO] {patch['file']}: target {patch['target']} not "
+                  f"present in this checkout; skipping")
+            continue
+        try:
+            text = target.read_text(encoding="utf-8", errors="ignore")
+        except OSError as exc:
+            print(f"[WARNING] {patch['file']}: could not read {target}: "
+                  f"{exc}; skipping")
+            continue
+        if re.search(patch["marker"], text):
+            print(f"[OK] {patch['file']}: fix already present, skipping")
+            continue
+        patch_path = PATCH_DIR / patch["file"]
+        if not patch_path.is_file():
+            print(f"[ERROR] {patch['file']}: patch file not found at "
+                  f"{patch_path}; cannot apply")
+            return 1
+        check_argv = ["git", "-C", str(audiocpp_dir), "apply", "--check",
+                      "--whitespace=nowarn", str(patch_path)]
+        check_rc = common.run_console_subprocess(
+            check_argv, emit=emit, cancel=cancel)
+        if check_rc != 0:
+            print(f"[ERROR] {patch['file']}: no longer applies to "
+                  f"{patch['target']} (git apply --check exit {check_rc}). "
+                  f"The audio.cpp fork's vendored ggml changed shape and "
+                  f"still lacks the fix. Re-evaluate {patch_path}: "
+                  f"regenerate the patch, or drop this entry if the fork "
+                  f"now ships the fix.")
+            return 1
+        apply_argv = ["git", "-C", str(audiocpp_dir), "apply",
+                      "--whitespace=nowarn", str(patch_path)]
+        rc = common.run_console_subprocess(
+            apply_argv, emit=emit, cancel=cancel)
+        if rc != 0:
+            print(f"[ERROR] {patch['file']}: git apply failed (exit {rc})")
+            return rc
+        print(f"[OK] {patch['file']}: applied ({patch['label']})")
+    return 0
+
+
 def build_audiocpp(audiocpp_dir: Path, backend: str, *,
                    emit=None, cancel=None) -> int:
     """Build audiocpp_server for BACKEND, streaming output.
@@ -1843,6 +1924,15 @@ def build_audiocpp(audiocpp_dir: Path, backend: str, *,
         if emit is not None:
             common.record_post_tui_notice(message)
         return 1
+    patch_rc = apply_ggml_patches(audiocpp_dir, emit=emit, cancel=cancel)
+    if patch_rc != 0:
+        message = ("[ERROR] ggml build patches could not be applied; "
+                   "aborting audiocpp_server build. See the messages above "
+                   "and re-evaluate app/backends/patches/.")
+        print(message)
+        if emit is not None:
+            common.record_post_tui_notice(message)
+        return patch_rc
     argv = ["sh", str(script), "--backend", backend, "--target",
             "audiocpp_server"]
     command = f"cd {audiocpp_dir} && {shlex.join(argv)}"
@@ -2159,6 +2249,13 @@ def _collect_from_flags(args: argparse.Namespace,
         if rc != 0:
             parser.error(f"git clone failed (exit {rc}); clone audio.cpp "
                          f"manually: git clone {AUDIOCPP_GIT_URL} {target}")
+        patch_rc = apply_ggml_patches(target)
+        if patch_rc != 0:
+            parser.error(
+                f"ggml build patches could not be applied to {target} "
+                f"(exit {patch_rc}); see messages above. The audio.cpp "
+                f"fork's vendored ggml may have changed — re-evaluate "
+                f"app/backends/patches/.")
         audiocpp_dir = target
     if audiocpp_dir is None:
         parser.error(
diff --git a/app/backends/patches/ggml-top-k-cuda-iterator.patch b/app/backends/patches/ggml-top-k-cuda-iterator.patch
new file mode 100644
index 0000000..0eb89a5
--- /dev/null
+++ b/app/backends/patches/ggml-top-k-cuda-iterator.patch
@@ -0,0 +1,10 @@
+--- a/external/ggml/src/ggml-cuda/top-k.cu
++++ b/external/ggml/src/ggml-cuda/top-k.cu
+@@ -4,6 +4,7 @@
+ #ifdef GGML_CUDA_USE_CUB
+ #    include <cub/cub.cuh>
+ #    if (CCCL_MAJOR_VERSION >= 3 && CCCL_MINOR_VERSION >= 2)
+ #        define CUB_TOP_K_AVAILABLE
++#        include <cuda/iterator>
+ using namespace cub;
+ #    endif  // CCCL_MAJOR_VERSION >= 3 && CCCL_MINOR_VERSION >= 2