Fix CRAN pretest failures: clang 21 errors/warnings + Windows byte ambiguity

Zachary DeBruine · Zachary DeBruine · commit 7809e9a44602 · 2026-03-14T12:45:16.000-04:00
- krylov.hpp: Remove erroneous .template on topRows(k) runtime calls
- dense_spz_loader.hpp, spz_loader.hpp: Fix member initializer order (-Wreorder)
- speckled_cv.hpp: Explicit cast uint64_t to double (-Wimplicit-const-int-float-conversion)
- spmv.hpp: Fix undefined 'n' variable, guard with #ifdef _OPENMP
- fused_nnls.hpp: Suppress unused 'threads' parameter in non-OpenMP paths
- fit_cv.hpp: Remove unused variables (actual_threads, t_conv_us)
- deflation.hpp: Remove dead k_computed assignment
- sparsepress_v2.hpp: Remove unused private fields, suppress num_threads warning
- fit_cpu.hpp: Fix %ld format with long long args on Windows (use int cast)
- platform.hpp, platform_io.hpp: Fix std::byte vs rpcndr.h byte ambiguity on MinGW
- Makevars.win: Add -Wno-ignored-attributes to suppress Eigen SSE noise
diff --git a/inst/build_config b/inst/build_config
@@ -4,4 +4,4 @@ cuda_version=
 cuda_home=
 nvcc_path=
 gpu_arch=
-build_date=2026-03-14T14:03:24Z
+build_date=2026-03-14T16:07:22Z
diff --git a/inst/include/FactorNet/core/platform.hpp b/inst/include/FactorNet/core/platform.hpp
@@ -16,7 +16,16 @@
 #include <string>
 
 #if defined(_WIN32)
+// Workaround for MinGW C++17+: std::byte (from <cstddef>) conflicts
+// with rpcndr.h "typedef unsigned char byte;" inside <windows.h>.
+// Temporarily rename the Windows byte to avoid the ambiguity.
+#define WIN32_LEAN_AND_MEAN
+#ifndef NOMINMAX
+#define NOMINMAX
+#endif
+#define byte win_byte_override
 #include <windows.h>
+#undef byte
 #elif defined(__APPLE__)
 #include <sys/sysctl.h>
 #include <unistd.h>
diff --git a/inst/include/FactorNet/io/dense_spz_loader.hpp b/inst/include/FactorNet/io/dense_spz_loader.hpp
@@ -52,7 +52,7 @@ class DenseSpzLoader : public DataLoader<Scalar> {
      */
     explicit DenseSpzLoader(const std::string& path,
                             bool require_transpose = true)
-        : fwd_idx_(0), trans_idx_(0), in_core_(false)
+        : in_core_(false), fwd_idx_(0), trans_idx_(0)
     {
         reader_ = std::make_unique<streampress::FileReader>(path);
         file_size_ = reader_->file_size();
diff --git a/inst/include/FactorNet/io/spz_loader.hpp b/inst/include/FactorNet/io/spz_loader.hpp
@@ -55,7 +55,7 @@ class SpzLoader : public DataLoader<Scalar> {
      * @param require_transpose If true (default), throw if file lacks transpose section
      */
     explicit SpzLoader(const std::string& path, bool require_transpose = true)
-        : fwd_idx_(0), trans_idx_(0), in_core_(false)
+        : in_core_(false), fwd_idx_(0), trans_idx_(0)
     {
         reader_ = std::make_unique<streampress::FileReader>(path);
         file_size_ = reader_->file_size();
@@ -303,7 +303,7 @@ class SpzLoader : public DataLoader<Scalar> {
             ? header_.metadata_offset
             : file_size_ - streampress::v2::FOOTER_SIZE;
 
-        uint64_t trans_section_size = trans_end - trans_start;
+        (void)(trans_end - trans_start);  // trans_section_size unused but validates range
 
         // Read the full file from offset 0 to trans_end into a buffer
         // so TransposeChunkReader can parse it (it needs header + transpose data)
diff --git a/inst/include/FactorNet/nmf/fit_cpu.hpp b/inst/include/FactorNet/nmf/fit_cpu.hpp
@@ -1813,9 +1813,9 @@ ::FactorNet::NMFResult<Scalar> nmf_fit(
         // --- Timer report ---
         _t_end = std::chrono::high_resolution_clock::now();
         if (config.verbose) {
-            auto us = [](auto a, auto b) { return std::chrono::duration_cast<std::chrono::microseconds>(b - a).count(); };
-            Rprintf("[iter %d] gram_h=%ldus rhs_h=%ldus nnls_h=%ldus norm_h=%ldus | "
-                    "gram_w=%ldus rhs_w=%ldus nnls_w=%ldus norm_w=%ldus | loss=%ldus | total=%ldus\n",
+            auto us = [](auto a, auto b) { return static_cast<int>(std::chrono::duration_cast<std::chrono::microseconds>(b - a).count()); };
+            Rprintf("[iter %d] gram_h=%dus rhs_h=%dus nnls_h=%dus norm_h=%dus | "
+                    "gram_w=%dus rhs_w=%dus nnls_w=%dus norm_w=%dus | loss=%dus | total=%dus\n",
                     iter, us(_t_gram_h, _t_rhs_h), us(_t_rhs_h, _t_nnls_h),
                     us(_t_nnls_h, _t_norm_h), us(_t_norm_h, _t_gram_w),
                     us(_t_gram_w, _t_rhs_w), us(_t_rhs_w, _t_nnls_w),
diff --git a/inst/include/FactorNet/nmf/fit_cv.hpp b/inst/include/FactorNet/nmf/fit_cv.hpp
@@ -364,16 +364,15 @@ ::FactorNet::NMFResult<Scalar> nmf_fit_cv(
         ? config.threads
         : omp_get_max_threads();
 #else
-    const int actual_threads = 1;
-#endif
     (void)n_threads;
+#endif
 
     // ------------------------------------------------------------------
     // 5. Iteration loop
     // ------------------------------------------------------------------
 
     // Timing accumulators (microseconds) for profiling
-    int64_t t_h_update_us = 0, t_w_update_us = 0, t_loss_us = 0, t_conv_us = 0;
+    int64_t t_h_update_us = 0, t_w_update_us = 0, t_loss_us = 0;
     auto tick = []() { return std::chrono::high_resolution_clock::now(); };
     auto elapsed_us = [](auto start, auto end) {
         return std::chrono::duration_cast<std::chrono::microseconds>(end - start).count();
diff --git a/inst/include/FactorNet/nmf/speckled_cv.hpp b/inst/include/FactorNet/nmf/speckled_cv.hpp
@@ -84,7 +84,7 @@ class LazySpeckledMask {
         h = (h ^ (h >> 30)) * 0xbf58476d1ce4e5b9ULL;
         h = (h ^ (h >> 27)) * 0x94d049bb133111ebULL;
         h ^= h >> 31;
-        uint64_t threshold = static_cast<uint64_t>(col_subsample_frac_ * 0xFFFFFFFFFFFFFFFFULL);
+        uint64_t threshold = static_cast<uint64_t>(col_subsample_frac_ * static_cast<double>(0xFFFFFFFFFFFFFFFFULL));
         return h < threshold;
     }
     
@@ -98,7 +98,7 @@ class LazySpeckledMask {
         h = (h ^ (h >> 30)) * 0xbf58476d1ce4e5b9ULL;
         h = (h ^ (h >> 27)) * 0x94d049bb133111ebULL;
         h ^= h >> 31;
-        uint64_t threshold = static_cast<uint64_t>(row_subsample_frac_ * 0xFFFFFFFFFFFFFFFFULL);
+        uint64_t threshold = static_cast<uint64_t>(row_subsample_frac_ * static_cast<double>(0xFFFFFFFFFFFFFFFFULL));
         return h < threshold;
     }
     
diff --git a/inst/include/FactorNet/primitives/cpu/fused_nnls.hpp b/inst/include/FactorNet/primitives/cpu/fused_nnls.hpp
@@ -93,7 +93,8 @@ inline void fused_rhs_nnls_sparse(
 #ifdef _OPENMP
     int n_threads = (threads > 0) ? threads : omp_get_max_threads();
 #else
-    int n_threads = 1;
+    int n_threads = 1;  // NOLINT(misc-const-correctness)
+    (void)threads;
 #endif
 
     const bool has_L1 = (L1 > Scalar(0));
@@ -175,6 +176,7 @@ inline void fused_rhs_cholesky_sparse(
     int n_threads = (threads > 0) ? threads : omp_get_max_threads();
 #else
     int n_threads = 1;
+    (void)threads;
 #endif
 
     // Pre-factorize G once — O(k³), amortized over n columns
diff --git a/inst/include/FactorNet/svd/deflation.hpp b/inst/include/FactorNet/svd/deflation.hpp
@@ -619,8 +619,6 @@ SVDResult<Scalar> deflation_svd(const MatrixType& A,
         col_weights.setOnes();
     }
 
-    int k_computed = 0;
-
     for (int k = 0; k < k_max; ++k) {
         SVD_CHECK_INTERRUPT();
 
@@ -865,7 +863,6 @@ SVDResult<Scalar> deflation_svd(const MatrixType& A,
         U_all.col(k) = u;
         d_all(k) = sigma;
         V_all.col(k) = v;
-        k_computed = k + 1;
 
         result.iters_per_factor.push_back(iter);
 
diff --git a/inst/include/FactorNet/svd/krylov.hpp b/inst/include/FactorNet/svd/krylov.hpp
@@ -228,12 +228,12 @@ void extract_svd_from_factors(
 
     DenseMatrix<Scalar> Q_w = qr_w.householderQ() *
         DenseMatrix<Scalar>::Identity(W.rows(), k);
-    DenseMatrix<Scalar> R_w = qr_w.matrixQR().template topRows(k)
+    DenseMatrix<Scalar> R_w = qr_w.matrixQR().topRows(k)
         .template triangularView<Eigen::Upper>();
 
     DenseMatrix<Scalar> Q_h = qr_h.householderQ() *
         DenseMatrix<Scalar>::Identity(H.rows(), k);
-    DenseMatrix<Scalar> R_h = qr_h.matrixQR().template topRows(k)
+    DenseMatrix<Scalar> R_h = qr_h.matrixQR().topRows(k)
         .template triangularView<Eigen::Upper>();
 
     // Small SVD of R_w · R_h'  (k × k)
diff --git a/inst/include/FactorNet/svd/spmv.hpp b/inst/include/FactorNet/svd/spmv.hpp
@@ -396,23 +396,23 @@ void spmv_forward_dense(const DenseMatrixType& A,
                         const DenseVector<Scalar>* row_inv_sds = nullptr)
 {
     const int m = static_cast<int>(A.rows());
-    const int n = static_cast<int>(A.cols());
     result.resize(m);
     result.setZero();
 
 #ifdef _OPENMP
     const int nthreads = threads > 0 ? threads : omp_get_max_threads();
     if (nthreads > 1) {
         // Column-scatter pattern (cache-friendly for column-major A)
-        const int nt = std::min(nthreads, n);
+        const int n_cols = static_cast<int>(A.cols());
+        const int nt = std::min(nthreads, n_cols);
         std::vector<DenseVector<Scalar>> locals(nt, DenseVector<Scalar>::Zero(m));
 
         #pragma omp parallel num_threads(nt)
         {
             const int tid = omp_get_thread_num();
             auto& local = locals[tid];
             #pragma omp for schedule(static)
-            for (int j = 0; j < n; ++j) {
+            for (int j = 0; j < n_cols; ++j) {
                 Scalar vj = v(j);
                 if (vj != static_cast<Scalar>(0))
                     local.noalias() += A.col(j) * vj;
@@ -527,23 +527,23 @@ void spmm_forward_dense(const DenseMatrixType& A,
                         const DenseVector<Scalar>* row_inv_sds = nullptr)
 {
     const int m = static_cast<int>(A.rows());
-    const int n = static_cast<int>(A.cols());
     const int l = static_cast<int>(X.cols());
     Y.resize(m, l);
     Y.setZero();
 
 #ifdef _OPENMP
     const int nthreads = threads > 0 ? threads : omp_get_max_threads();
     if (nthreads > 1) {
-        const int nt = std::min(nthreads, n);
+        const int n_cols = static_cast<int>(A.cols());
+        const int nt = std::min(nthreads, n_cols);
         std::vector<DenseMatrix<Scalar>> locals(nt, DenseMatrix<Scalar>::Zero(m, l));
 
         #pragma omp parallel num_threads(nt)
         {
             const int tid = omp_get_thread_num();
             auto& local = locals[tid];
             #pragma omp for schedule(static)
-            for (int j = 0; j < n; ++j) {
+            for (int j = 0; j < n_cols; ++j) {
                 for (int c = 0; c < l; ++c) {
                     local.col(c).noalias() += A.col(j) * X(j, c);
                 }
diff --git a/inst/include/streampress/core/platform_io.hpp b/inst/include/streampress/core/platform_io.hpp
@@ -12,7 +12,14 @@
 #include <cstddef>
 
 #ifdef _WIN32
+// Workaround for MinGW C++17+: std::byte vs rpcndr.h byte collision
+#define WIN32_LEAN_AND_MEAN
+#ifndef NOMINMAX
+#define NOMINMAX
+#endif
+#define byte win_byte_override
 #include <windows.h>
+#undef byte
 #include <io.h>
 #else
 #include <unistd.h>
diff --git a/inst/include/streampress/sparsepress_v2.hpp b/inst/include/streampress/sparsepress_v2.hpp
@@ -962,8 +962,9 @@ inline CSCMatrix decompress_v2(
         chunk_nnz_offset[c + 1] = chunk_nnz_offset[c] + chunk_descs[first_chunk + c].nnz;
 
     int num_threads = cfg.num_threads;
+    (void)num_threads;
     #ifdef _OPENMP
-    if (num_threads <= 0) num_threads = omp_get_max_threads();
+    num_threads = (num_threads <= 0) ? omp_get_max_threads() : num_threads;
     // Clamp to actual chunk count: idle threads beyond n_decode cause race
     // conditions in repeated calls (e.g., loading many files in a loop).
     num_threads = std::min(num_threads, static_cast<int>(n_decode));
@@ -1490,7 +1491,6 @@ inline CSCMatrix decompress_v2_transpose(
 class TransposeChunkReader {
 public:
     TransposeChunkReader(const uint8_t* data, size_t data_size)
-        : data_(data), data_size_(data_size)
     {
         if (data_size < HEADER_SIZE_V2)
             throw std::runtime_error("Data too small for v2 header");
@@ -1645,8 +1645,6 @@ class TransposeChunkReader {
     }
 
 private:
-    const uint8_t* data_;
-    size_t data_size_;
     FileHeader_v2 hdr_;
     ValueType_v2 vtype_;
     uint32_t num_t_chunks_;
diff --git a/src/Makevars.win b/src/Makevars.win
@@ -2,6 +2,6 @@ PKG_CPPFLAGS = -I../inst/include/
 PKG_LIBS = $(SHLIB_OPENMP_CXXFLAGS) $(LAPACK_LIBS) $(BLAS_LIBS) $(FLIBS)
 
 ## Compiler flags for maximum performance
-## -O3: Aggressive optimization including auto-vectorization
 ## -Wa,-mbig-obj: Allow large object files (needed for heavily templated Eigen code on Windows)
-PKG_CXXFLAGS = $(SHLIB_OPENMP_CXXFLAGS) -DEIGEN_INITIALIZE_MATRICES_BY_ZERO -DEIGEN_NO_DEBUG -DSPARSEPRESS_USE_R -DFACTORNET_USE_R -Wa,-mbig-obj
+## -Wno-ignored-attributes: Suppress harmless Eigen SSE/AVX type-attribute warnings
+PKG_CXXFLAGS = $(SHLIB_OPENMP_CXXFLAGS) -DEIGEN_INITIALIZE_MATRICES_BY_ZERO -DEIGEN_NO_DEBUG -DSPARSEPRESS_USE_R -DFACTORNET_USE_R -Wa,-mbig-obj -Wno-ignored-attributes
diff --git a/tools/cran_submit.R b/tools/cran_submit.R
@@ -0,0 +1,75 @@
+#!/usr/bin/env Rscript
+# Non-interactive CRAN submission script
+# Uploads tarball, fills Step 2 form, triggers confirmation email
+
+built_path <- "/mnt/home/debruinz/RcppML-2/RcppML_1.0.0.tar.gz"
+stopifnot(file.exists(built_path))
+cat("Tarball:", built_path, "\n")
+cat("Size:", round(file.size(built_path) / 1e6, 1), "MB\n")
+
+comment <- paste(readLines("/mnt/home/debruinz/RcppML-2/cran-comments.md"), collapse = "\n")
+
+# Step 1: Upload the tarball
+cat("\n=== Step 1: Uploading tarball ===\n")
+h1 <- curl::new_handle()
+curl::handle_setopt(h1, cookiejar = "", cookiefile = "",
+                    timeout = 300, connecttimeout = 60,
+                    low_speed_time = 300, low_speed_limit = 100)
+curl::handle_setform(h1,
+  pkg_id = "",
+  name = "Zachary DeBruine",
+  email = "zacharydebruine@gmail.com",
+  uploaded_file = curl::form_file(built_path, type = "application/x-gzip"),
+  upload = "Upload the package"
+)
+resp1 <- curl::curl_fetch_memory("https://xmpalantir.wu.ac.at/cransubmit/index2.php", handle = h1)
+page1 <- rawToChar(resp1$content)
+cat("HTTP status:", resp1$status_code, "\n")
+
+# Extract pkg_id from the step 2 form
+pkg_id_match <- regmatches(page1, regexpr('name="pkg_id"[^>]*value="[^"]+"', page1))
+if (length(pkg_id_match) == 0) {
+  cat("ERROR: Could not find pkg_id in response. Dumping page:\n")
+  cat(page1)
+  stop("Upload failed - no pkg_id found")
+}
+pkg_id <- sub('.*value="', '', sub('"$', '', pkg_id_match))
+cat("pkg_id:", pkg_id, "\n")
+
+# Extract all hidden inputs and form fields
+inputs <- regmatches(page1, gregexpr("<input[^>]+>", page1))[[1]]
+cat("\nForm inputs found:\n")
+for (inp in inputs) cat("  ", inp, "\n")
+
+# Step 2: Submit the form with comment
+cat("\n=== Step 2: Submitting to CRAN ===\n")
+h2 <- curl::new_handle()
+curl::handle_setopt(h2, cookiejar = "", cookiefile = "", timeout = 120)
+# Copy cookies from h1
+cookies <- curl::handle_cookies(h1)
+cat("Cookies:", nrow(cookies), "\n")
+
+curl::handle_setform(h2,
+  pkg_id = pkg_id,
+  name = "Zachary DeBruine",
+  email = "zacharydebruine@gmail.com",
+  comment = comment,
+  submit = "Submit package"
+)
+resp2 <- curl::curl_fetch_memory("https://xmpalantir.wu.ac.at/cransubmit/index2.php", handle = h2)
+page2 <- rawToChar(resp2$content)
+cat("HTTP status:", resp2$status_code, "\n")
+
+# Check result
+if (grepl("confirm", page2, ignore.case = TRUE)) {
+  cat("\n=== SUCCESS: Package submitted! ===\n")
+  cat("A confirmation email has been sent to zacharydebruine@gmail.com\n")
+  cat("You MUST click the link in that email to complete the submission.\n")
+} else {
+  cat("\nResponse page content:\n")
+  # Strip HTML tags for readability
+  lines <- strsplit(page2, "<[^>]+>")[[1]]
+  lines <- trimws(lines)
+  lines <- lines[nchar(lines) > 3]
+  cat(lines, sep = "\n")
+}

Original file line number	Diff line number	Diff line change
`@@ -52,7 +52,7 @@ class DenseSpzLoader : public DataLoader<Scalar> {`
`52`	`52`	`*/`
`53`	`53`	`explicit DenseSpzLoader(const std::string& path,`
`54`	`54`	`bool require_transpose = true)`
`55`		`- : fwd_idx_(0), trans_idx_(0), in_core_(false)`
	`55`	`+ : in_core_(false), fwd_idx_(0), trans_idx_(0)`
`56`	`56`	`{`
`57`	`57`	`reader_ = std::make_unique<streampress::FileReader>(path);`
`58`	`58`	`file_size_ = reader_->file_size();`