tests compile but not link

external/CMakeLists.txt

@@ -64,19 +64,25 @@ elseif(EXISTS "${liblinear_SOURCE_DIR}/newton.cpp")
     list(APPEND LIBLINEAR_SOURCES "${liblinear_SOURCE_DIR}/newton.cpp")
 endif()
 
-# Check for BLAS files in blas directory
+# Check for BLAS files in blas directory - Fixed to ensure they're included
-file(GLOB BLAS_C_FILES "${liblinear_SOURCE_DIR}/blas/*.c")
+if(EXISTS "${liblinear_SOURCE_DIR}/blas")
-file(GLOB BLAS_CPP_FILES "${liblinear_SOURCE_DIR}/blas/*.cpp")
+    # Explicitly add the required BLAS source files
+    set(BLAS_FILES
+        "${liblinear_SOURCE_DIR}/blas/daxpy.c"
+        "${liblinear_SOURCE_DIR}/blas/ddot.c"
+        "${liblinear_SOURCE_DIR}/blas/dnrm2.c"
+        "${liblinear_SOURCE_DIR}/blas/dscal.c"
+    )
     
-if(BLAS_C_FILES OR BLAS_CPP_FILES)
+    # Check which BLAS files actually exist and add them
-    list(APPEND LIBLINEAR_SOURCES ${BLAS_C_FILES} ${BLAS_CPP_FILES})
+    foreach(blas_file ${BLAS_FILES})
-else()
+        if(EXISTS ${blas_file})
-    # Try alternative BLAS file names
+            list(APPEND LIBLINEAR_SOURCES ${blas_file})
-    foreach(blas_file daxpy ddot dnrm2 dscal)
+            message(STATUS "Adding BLAS file: ${blas_file}")
-        if(EXISTS "${liblinear_SOURCE_DIR}/blas/${blas_file}.c")
-            list(APPEND LIBLINEAR_SOURCES "${liblinear_SOURCE_DIR}/blas/${blas_file}.c")
         endif()
     endforeach()
+else()
+    message(WARNING "BLAS directory not found in liblinear source")
 endif()
 
 # Create liblinear object library if we have source files

tests/test_multiclass_strategy.cpp

@@ -16,9 +16,10 @@ using namespace svm_classifier;
  */
 std::pair<torch::Tensor, torch::Tensor> generate_multiclass_data(int n_samples = 60,
     int n_features = 2,
-    int n_classes = 3)
+    int n_classes = 3,
+    int seed = 42)
 {
-    torch::manual_seed(42);
+    torch::manual_seed(seed);
 
     auto X = torch::randn({ n_samples, n_features });
     auto y = torch::randint(0, n_classes, { n_samples });
@@ -123,7 +124,7 @@ TEST_CASE("OneVsRestStrategy Prediction", "[unit][multiclass_strategy]")
     {
         auto predictions = strategy.predict(X_test, converter);
 
-        REQUIRE(predictions.size() == X_test.size(0));
+        REQUIRE(static_cast<int64_t>(predictions.size()) == X_test.size(0));
 
         // Check that all predictions are valid class labels
         auto classes = strategy.get_classes();
@@ -136,8 +137,8 @@ TEST_CASE("OneVsRestStrategy Prediction", "[unit][multiclass_strategy]")
     {
         auto decision_values = strategy.decision_function(X_test, converter);
 
-        REQUIRE(decision_values.size() == X_test.size(0));
+        REQUIRE(static_cast<int64_t>(decision_values.size()) == X_test.size(0));
-        REQUIRE(decision_values[0].size() == strategy.get_n_classes());
+        REQUIRE(static_cast<int>(decision_values[0].size()) == strategy.get_n_classes());
 
         // Decision values should be real numbers
         for (const auto& sample_decisions : decision_values) {
@@ -173,7 +174,7 @@ TEST_CASE("OneVsRestStrategy Probability Prediction", "[unit][multiclass_strateg
         if (strategy.supports_probability()) {
             auto probabilities = strategy.predict_proba(X, converter);
 
-            REQUIRE(probabilities.size() == X.size(0));
+            REQUIRE(static_cast<int64_t>(probabilities.size()) == X.size(0));
             REQUIRE(probabilities[0].size() == 3);  // 3 classes
 
             // Check probability constraints
@@ -269,7 +270,7 @@ TEST_CASE("OneVsOneStrategy Prediction", "[unit][multiclass_strategy]")
     {
         auto predictions = strategy.predict(X_test, converter);
 
-        REQUIRE(predictions.size() == X_test.size(0));
+        REQUIRE(static_cast<int64_t>(predictions.size()) == X_test.size(0));
 
         auto classes = strategy.get_classes();
         for (int pred : predictions) {
@@ -281,7 +282,7 @@ TEST_CASE("OneVsOneStrategy Prediction", "[unit][multiclass_strategy]")
     {
         auto decision_values = strategy.decision_function(X_test, converter);
 
-        REQUIRE(decision_values.size() == X_test.size(0));
+        REQUIRE(static_cast<int64_t>(decision_values.size()) == X_test.size(0));
 
         // For 3 classes, OvO should have C(3,2) = 3 pairwise comparisons
         REQUIRE(decision_values[0].size() == 3);
@@ -298,7 +299,7 @@ TEST_CASE("OneVsOneStrategy Prediction", "[unit][multiclass_strategy]")
         // OvO probability estimation is more complex
         auto probabilities = strategy.predict_proba(X_test, converter);
 
-        REQUIRE(probabilities.size() == X_test.size(0));
+        REQUIRE(static_cast<int64_t>(probabilities.size()) == X_test.size(0));
         REQUIRE(probabilities[0].size() == 3);  // 3 classes
 
         // Check basic probability constraints
@@ -395,7 +396,7 @@ TEST_CASE("MulticlassStrategy Edge Cases", "[unit][multiclass_strategy]")
         // Implementation might extend to binary case
 
         auto predictions = strategy.predict(X, converter);
-        REQUIRE(predictions.size() == X.size(0));
+        REQUIRE(static_cast<int64_t>(predictions.size()) == X.size(0));
     }
 
     SECTION("Very small dataset")
@@ -432,7 +433,7 @@ TEST_CASE("MulticlassStrategy Edge Cases", "[unit][multiclass_strategy]")
         REQUIRE(strategy.get_n_classes() == 2);
 
         auto predictions = strategy.predict(X, converter);
-        REQUIRE(predictions.size() == X.size(0));
+        REQUIRE(static_cast<int64_t>(predictions.size()) == X.size(0));
     }
 }
 
@@ -510,7 +511,7 @@ TEST_CASE("MulticlassStrategy Memory Management", "[unit][multiclass_strategy]")
             REQUIRE(metrics.status == TrainingStatus::SUCCESS);
 
             auto predictions = strategy.predict(X, converter);
-            REQUIRE(predictions.size() == X.size(0));
+            REQUIRE(static_cast<int64_t>(predictions.size()) == X.size(0));
         }
     }
 }