Suppress non-const global variable warnings and fix security integer casts (#5092)

Summary:
Pull Request resolved: #5092

Fix 32 `facebook-avoid-non-const-global-variables` warnings by adding NOLINT comments to intentionally mutable global stats objects and configuration variables that are part of the public API (e.g., `indexIVF_stats`, `hnsw_stats`, `check_compatible_for_merge_expensive_check`). These cannot be made const as they accumulate runtime statistics and provide runtime configuration.

Fix 25 `facebook-security-vulnerable-integer-implicit-cast` warnings by adding explicit `static_cast<size_t>()` to one operand before integer multiplication to prevent overflow before widening. Also fix `clang-diagnostic-switch-enum` warnings by adding missing enum cases to switch statements, and various other minor lint fixes including NOLINT for `-inl.h` includes, `ShadowingClass` fix, and `BadImplicitCast` fix.

Reviewed By: mnorris11

Differential Revision: D100592783

Author: junjieqi

faiss/Index2Layer.cpp

@@ -266,7 +266,8 @@ DistanceComputer* Index2Layer::get_distance_computer() const {
 /* The standalone codec interface */
 
 // block size used in Index2Layer::sa_encode
-int index2layer_sa_encode_bs = 32768;
+int index2layer_sa_encode_bs =
+        32768; // NOLINT(facebook-avoid-non-const-global-variables)
 
 void Index2Layer::sa_encode(idx_t n, const float* x, uint8_t* bytes) const {
     FAISS_THROW_IF_NOT(is_trained);

faiss/IndexAdditiveQuantizer.cpp

@@ -47,7 +47,7 @@ struct AQDistanceComputerDecompress : FlatCodesDistanceComputer {
             const IndexAdditiveQuantizer& iaq,
             VectorDistance vd_)
             : FlatCodesDistanceComputer(iaq.codes.data(), iaq.code_size),
-              tmp(iaq.d * 2),
+              tmp(static_cast<size_t>(iaq.d) * 2),
               aq(*iaq.aq),
               vd(vd_),
               d(iaq.d) {}
@@ -79,7 +79,7 @@ struct AQDistanceComputerLUT : FlatCodesDistanceComputer {
 
     explicit AQDistanceComputerLUT(const IndexAdditiveQuantizer& iaq)
             : FlatCodesDistanceComputer(iaq.codes.data(), iaq.code_size),
-              LUT(iaq.aq->total_codebook_size + iaq.d * 2),
+              LUT(iaq.aq->total_codebook_size + static_cast<size_t>(iaq.d) * 2),
               aq(*iaq.aq),
               d(iaq.d) {}
 
@@ -219,6 +219,9 @@ FlatCodesDistanceComputer* IndexAdditiveQuantizer::
                             false,
                             AdditiveQuantizer::ST_norm_cqint8>(*this);
 #undef DISPATCH
+                case AdditiveQuantizer::ST_decompress:
+                case AdditiveQuantizer::ST_norm_from_LUT:
+                case AdditiveQuantizer::ST_count:
                 default:
                     FAISS_THROW_FMT(
                             "search type %d not supported", aq->search_type);
@@ -276,6 +279,8 @@ void IndexAdditiveQuantizer::search(
                             *this, x, rh);
                     break;
 #undef DISPATCH
+                case AdditiveQuantizer::ST_decompress:
+                case AdditiveQuantizer::ST_count:
                 default:
                     FAISS_THROW_FMT(
                             "search type %d not supported", aq->search_type);

faiss/IndexBinaryHash.cpp

@@ -273,7 +273,8 @@ void IndexBinaryHashStats::reset() {
     memset((void*)this, 0, sizeof(*this));
 }
 
-IndexBinaryHashStats indexBinaryHash_stats;
+IndexBinaryHashStats
+        indexBinaryHash_stats; // NOLINT(facebook-avoid-non-const-global-variables)
 
 /*******************************************************
  * IndexBinaryMultiHash implementation

faiss/IndexFastScan.cpp

@@ -515,7 +515,8 @@ void IndexFastScan::search_implem_12(
     }
 }
 
-FastScanStats FastScan_stats;
+FastScanStats
+        FastScan_stats; // NOLINT(facebook-avoid-non-const-global-variables)
 
 template <class C>
 void IndexFastScan::search_implem_14(

faiss/IndexFlat.cpp

@@ -68,7 +68,7 @@ void IndexFlat::range_search(
     FAISS_THROW_IF_NOT_MSG(result, "RangeSearchResult object must not be null");
     IDSelector* sel = params ? params->sel : nullptr;
 
-    switch (metric_type) {
+    switch (metric_type) { // NOLINT(clang-diagnostic-switch-enum)
         case METRIC_INNER_PRODUCT:
             range_search_inner_product(
                     x, get_xb(), d, n, ntotal, radius, result, sel);
@@ -88,7 +88,7 @@ void IndexFlat::compute_distance_subset(
         float* distances,
         const idx_t* labels) const {
     FAISS_THROW_IF_NOT(k > 0);
-    switch (metric_type) {
+    switch (metric_type) { // NOLINT(clang-diagnostic-switch-enum)
         case METRIC_INNER_PRODUCT:
             fvec_inner_products_by_idx(distances, x, get_xb(), labels, d, n, k);
             break;

faiss/IndexHNSW.cpp

@@ -37,7 +37,7 @@ namespace faiss {
 using storage_idx_t = HNSW::storage_idx_t;
 using NodeDistFarther = HNSW::NodeDistFarther;
 
-HNSWStats hnsw_stats;
+HNSWStats hnsw_stats; // NOLINT(facebook-avoid-non-const-global-variables)
 
 /**************************************************************
  * add / search blocks of descriptors

faiss/IndexIVF.cpp

@@ -17,7 +17,6 @@
 #include <cstdio>
 #include <limits>
 
-#include <faiss/utils/hamming.h>
 #include <faiss/utils/utils.h>
 
 #include <faiss/IndexFlat.h>
@@ -1276,7 +1275,8 @@ void IndexIVF::train_encoder(
     }
 }
 
-bool check_compatible_for_merge_expensive_check = true;
+bool check_compatible_for_merge_expensive_check =
+        true; // NOLINT(facebook-avoid-non-const-global-variables)
 
 void IndexIVF::check_compatible_for_merge(const Index& otherIndex) const {
     // minimal sanity checks
@@ -1365,7 +1365,8 @@ void IndexIVFStats::add(const IndexIVFStats& other) {
     search_time += other.search_time;
 }
 
-IndexIVFStats indexIVF_stats;
+IndexIVFStats
+        indexIVF_stats; // NOLINT(facebook-avoid-non-const-global-variables)
 
 /*************************************************************************
  * InvertedListScanner

faiss/IndexIVFFastScan.cpp

@@ -1529,6 +1529,7 @@ void IndexIVFFastScan::postprocess_packed_codes(
         size_t /*n_added*/,
         const uint8_t* /*flat_codes*/) {}
 
-IVFFastScanStats IVFFastScan_stats;
+IVFFastScanStats
+        IVFFastScan_stats; // NOLINT(facebook-avoid-non-const-global-variables)
 
 } // namespace faiss

faiss/IndexIVFPQ.cpp

@@ -28,7 +28,7 @@
 #include <faiss/impl/IDSelector.h>
 #include <faiss/impl/ProductQuantizer.h>
 #include <faiss/impl/ResultHandler.h>
-#include <faiss/impl/pq_code_distance/pq_code_distance-inl.h>
+#include <faiss/impl/pq_code_distance/pq_code_distance-inl.h> // NOLINT(facebook-hte-InlineHeader)
 #include <faiss/impl/simd_dispatch.h>
 
 namespace faiss {
@@ -222,7 +222,8 @@ void IndexIVFPQ::sa_decode(idx_t n, const uint8_t* codes, float* x) const {
 }
 
 // block size used in IndexIVFPQ::add_core_o
-int index_ivfpq_add_core_o_bs = 32768;
+int index_ivfpq_add_core_o_bs =
+        32768; // NOLINT(facebook-avoid-non-const-global-variables)
 
 void IndexIVFPQ::add_core_o(
         idx_t n,
@@ -349,7 +350,8 @@ void IndexIVFPQ::reconstruct_from_offset(
 }
 
 /// 2G by default, accommodates tables up to PQ32 w/ 65536 centroids
-size_t precomputed_table_max_bytes = ((size_t)1) << 31;
+size_t precomputed_table_max_bytes = ((size_t)1)
+        << 31; // NOLINT(facebook-avoid-non-const-global-variables)
 
 /** Precomputed tables for residuals
  *
@@ -1301,10 +1303,14 @@ InvertedListScanner* IndexIVFPQ::get_InvertedListScanner(
     });
 }
 
-IndexIVFPQStats indexIVFPQ_stats;
+IndexIVFPQStats
+        indexIVFPQ_stats; // NOLINT(facebook-avoid-non-const-global-variables)
 
 void IndexIVFPQStats::reset() {
-    memset(this, 0, sizeof(*this));
+    nrefine = 0;
+    n_hamming_pass = 0;
+    search_cycles = 0;
+    refine_cycles = 0;
 }
 
 IndexIVFPQ::IndexIVFPQ() {

faiss/IndexIVFRaBitQ.cpp

@@ -332,6 +332,8 @@ void IndexIVFRaBitQ::sa_decode(idx_t n, const uint8_t* codes, float* x) const {
     }
 }
 
+namespace {
+
 struct IVFRaBitDistanceComputer : DistanceComputer {
     const float* q = nullptr;
     const IndexIVFRaBitQ* parent = nullptr;
@@ -379,6 +381,8 @@ float IVFRaBitDistanceComputer::symmetric_dis(idx_t /*i*/, idx_t /*j*/) {
     FAISS_THROW_MSG("Not implemented");
 }
 
+} // anonymous namespace
+
 DistanceComputer* IndexIVFRaBitQ::get_distance_computer() const {
     IVFRaBitDistanceComputer* dc = new IVFRaBitDistanceComputer;
     dc->parent = this;