#Centipede Extend the seed corpus maker library to support legacy inp…

…ut globs.

This is the easiest way to support seeding regression/crash inputs for replaying. This feature may also be useful in other contexts.

PiperOrigin-RevId: 703558948

centipede/BUILD

@@ -1122,6 +1122,7 @@ cc_library(
         ":thread_pool",
         ":util",
         ":workdir",
+        "@com_google_absl//absl/container:flat_hash_set",
         "@com_google_absl//absl/log",
         "@com_google_absl//absl/log:check",
         "@com_google_absl//absl/random",
@@ -1689,8 +1690,11 @@ cc_test(
         ":feature",
         ":seed_corpus_maker_lib",
         ":workdir",
+        "@com_google_absl//absl/log:check",
         "@com_google_absl//absl/strings",
+        "@com_google_fuzztest//common:defs",
         "@com_google_fuzztest//common:logging",
+        "@com_google_fuzztest//common:remote_file",
         "@com_google_fuzztest//common:test_util",
         "@com_google_googletest//:gtest_main",
     ],

centipede/seed_corpus_maker_lib.cc

@@ -36,6 +36,7 @@
 #include <variant>
 #include <vector>
 
+#include "absl/container/flat_hash_set.h"
 #include "absl/log/check.h"
 #include "absl/log/log.h"
 #include "absl/random/random.h"
@@ -115,21 +116,54 @@ absl::Status SampleSeedCorpusElementsFromSource(  //
     LOG(INFO) << "Selected " << src_dirs.size() << " corpus dir(s)";
   }
 
-  // Find all the corpus shard files in the found dirs.
+  // Find all the corpus shard and individual input files in the found dirs.
 
   std::vector<std::string> corpus_shard_fnames;
+  std::vector<std::string> individual_input_fnames;
   for (const auto& dir : src_dirs) {
-    const std::string shards_glob = fs::path{dir} / source.shard_rel_glob;
-    // NOTE: `RemoteGlobMatch` appends to the output list.
-    const auto prev_num_shards = corpus_shard_fnames.size();
-    RETURN_IF_NOT_OK(RemoteGlobMatch(shards_glob, corpus_shard_fnames));
-    LOG(INFO) << "Found " << (corpus_shard_fnames.size() - prev_num_shards)
-              << " shard(s) matching " << shards_glob;
+    absl::flat_hash_set<std::string> current_corpus_shard_fnames;
+    if (!source.shard_rel_glob.empty()) {
+      std::vector<std::string> matched_fnames;
+      const std::string glob = fs::path{dir} / source.shard_rel_glob;
+      const auto match_status = RemoteGlobMatch(glob, matched_fnames);
+      if (!match_status.ok() && !absl::IsNotFound(match_status)) {
+        LOG(ERROR) << "Got error when glob-matching in " << dir << ": "
+                   << match_status;
+      } else {
+        current_corpus_shard_fnames.insert(matched_fnames.begin(),
+                                           matched_fnames.end());
+        corpus_shard_fnames.insert(corpus_shard_fnames.end(),
+                                   matched_fnames.begin(),
+                                   matched_fnames.end());
+        LOG(INFO) << "Found " << matched_fnames.size() << " shard(s) matching "
+                  << glob;
+      }
+    }
+    if (!source.individual_input_rel_glob.empty()) {
+      std::vector<std::string> matched_fnames;
+      const std::string glob = fs::path{dir} / source.individual_input_rel_glob;
+      const auto match_status = RemoteGlobMatch(glob, matched_fnames);
+      if (!match_status.ok() && !absl::IsNotFound(match_status)) {
+        LOG(ERROR) << "Got error when glob-matching in " << dir << ": "
+                   << match_status;
+      } else {
+        size_t num_added_individual_inputs = 0;
+        for (auto& fname : matched_fnames) {
+          if (current_corpus_shard_fnames.contains(fname)) continue;
+          if (RemotePathIsDirectory(fname)) continue;
+          ++num_added_individual_inputs;
+          individual_input_fnames.push_back(std::move(fname));
+        }
+        LOG(INFO) << "Found " << num_added_individual_inputs
+                  << " individual input(s) with glob " << glob;
+      }
+    }
   }
-  LOG(INFO) << "Found " << corpus_shard_fnames.size()
-            << " shard(s) total in source " << source.dir_glob;
+  LOG(INFO) << "Found " << corpus_shard_fnames.size() << " shard(s) and "
+            << individual_input_fnames.size()
+            << " individual input(s) total in source " << source.dir_glob;
 
-  if (corpus_shard_fnames.empty()) {
+  if (corpus_shard_fnames.empty() && individual_input_fnames.empty()) {
     LOG(WARNING) << "Skipping empty source " << source.dir_glob;
     return absl::OkStatus();
   }
@@ -140,10 +174,12 @@ absl::Status SampleSeedCorpusElementsFromSource(  //
   const auto num_shards = corpus_shard_fnames.size();
   std::vector<InputAndFeaturesVec> src_elts_per_shard(num_shards);
   std::vector<size_t> src_elts_with_features_per_shard(num_shards, 0);
+  InputAndFeaturesVec src_elts;
 
   {
     constexpr int kMaxReadThreads = 32;
-    ThreadPool threads{std::min<int>(kMaxReadThreads, num_shards)};
+    ThreadPool threads{std::min<int>(
+        kMaxReadThreads, std::max(num_shards, individual_input_fnames.size()))};
 
     for (int shard = 0; shard < num_shards; ++shard) {
       const auto& corpus_fname = corpus_shard_fnames[shard];
@@ -193,11 +229,27 @@ absl::Status SampleSeedCorpusElementsFromSource(  //
 
       threads.Schedule(read_shard);
     }
+
+    RPROF_SNAPSHOT_AND_LOG("Done reading shards");
+
+    src_elts.resize(individual_input_fnames.size());
+    for (size_t index = 0; index < individual_input_fnames.size(); ++index) {
+      threads.Schedule([index, &individual_input_fnames, &src_elts] {
+        ByteArray input;
+        const auto& path = individual_input_fnames[index];
+        const auto read_status = RemoteFileGetContents(path, input);
+        if (!read_status.ok()) {
+          LOG(WARNING) << "Skipping individual input path " << path
+                       << " due to read error: " << read_status;
+          return;
+        }
+        src_elts[index] = {std::move(input), {}};
+      });
+    }
   }
 
   RPROF_SNAPSHOT_AND_LOG("Done reading");
 
-  InputAndFeaturesVec src_elts;
   size_t src_num_features = 0;
 
   for (int s = 0; s < num_shards; ++s) {
@@ -217,6 +269,16 @@ absl::Status SampleSeedCorpusElementsFromSource(  //
 
   RPROF_SNAPSHOT_AND_LOG("Done merging");
 
+  // Remove empty inputs possibly due to read errors.
+  auto remove_it =
+      std::remove_if(src_elts.begin(), src_elts.end(),
+                     [](const auto& elt) { return std::get<0>(elt).empty(); });
+  if (remove_it != src_elts.end()) {
+    LOG(WARNING) << "Removed " << std::distance(remove_it, src_elts.end())
+                 << " empty inputs";
+    src_elts.erase(remove_it, src_elts.end());
+  }
+
   LOG(INFO) << "Read total of " << src_elts.size() << " elements ("
             << src_num_features << " with features) from source "
             << source.dir_glob;

centipede/seed_corpus_maker_lib.h

@@ -33,11 +33,16 @@ namespace centipede {
 // Native struct used by the seed corpus library for seed corpus source.
 //
 // TODO(b/362576261): Currently this is mirroring the `proto::SeedCorpusSource`
-// proto. But in the future it may change with the core seeding API.
+// proto. But in the future it may change with the core seeding API - any
+// difference is commented below.
 struct SeedCorpusSource {
   std::string dir_glob;
   uint32_t num_recent_dirs;
   std::string shard_rel_glob;
+  // If non-empty, will be used to glob the individual input files (with one
+  // input in each file) in the source dirs. Any files matching `shard_rel_glob`
+  // will be skipped.
+  std::string individual_input_rel_glob;
   std::variant<float, uint32_t> sampled_fraction_or_count;
 };
 

centipede/seed_corpus_maker_lib_test.cc

@@ -21,13 +21,17 @@
 #include <filesystem>  // NOLINT
 #include <string>
 #include <string_view>
+#include <vector>
 
 #include "gmock/gmock.h"
 #include "gtest/gtest.h"
+#include "absl/log/check.h"
 #include "absl/strings/str_cat.h"
 #include "./centipede/feature.h"
 #include "./centipede/workdir.h"
+#include "./common/defs.h"
 #include "./common/logging.h"  // IWYU pragma: keep
+#include "./common/remote_file.h"
 #include "./common/test_util.h"
 
 namespace centipede {
@@ -36,6 +40,7 @@ namespace {
 namespace fs = std::filesystem;
 
 using ::testing::IsSubsetOf;
+using ::testing::IsSupersetOf;
 
 inline constexpr auto kIdxDigits = WorkDir::kDigitsInShardIndex;
 
@@ -178,5 +183,69 @@ TEST(SeedCorpusMakerLibTest, RoundTripWriteReadWrite) {
   }
 }
 
+TEST(SeedCorpusMakerLibTest, LoadsBothIndividualInputsAndShardsFromSource) {
+  const fs::path test_dir = GetTestTempDir(test_info_->name());
+  chdir(test_dir.c_str());
+
+  const InputAndFeaturesVec kShardedInputs = {
+      {{0}, {}},
+      {{1}, {feature_domains::kNoFeature}},
+      {{0, 1}, {0x11, 0x23}},
+  };
+  constexpr std::string_view kCovBin = "bin";
+  constexpr std::string_view kCovHash = "hash";
+  constexpr std::string_view kRelDir = "dir/foo";
+
+  const std::vector<ByteArray> kIndividualInputs = {
+      {0, 1, 2},
+      {0, 1, 2, 3},
+      // Empty input expected to be not in the sample result.
+      {}};
+  // Write sharded inputs.
+  {
+    constexpr size_t kNumShards = 2;
+    const SeedCorpusDestination destination = {
+        .dir_path = std::string(kRelDir),
+        .shard_rel_glob = absl::StrCat("distilled-", kCovBin, ".*"),
+        .shard_index_digits = kIdxDigits,
+        .num_shards = kNumShards,
+    };
+    CHECK_OK(WriteSeedCorpusElementsToDestination(  //
+        kShardedInputs, kCovBin, kCovHash, destination));
+    const std::string workdir = (test_dir / kRelDir).c_str();
+    ASSERT_NO_FATAL_FAILURE(VerifyShardsExist(  //
+        workdir, kCovBin, kCovHash, kNumShards, ShardType::kDistilled));
+  }
+
+  // Write individual inputs
+  for (int i = 0; i < kIndividualInputs.size(); ++i) {
+    const auto path = std::filesystem::path(test_dir) / kRelDir /
+                      absl::StrCat("individual_input_", i);
+    CHECK_OK(RemoteFileSetContents(path.string(), kIndividualInputs[i]));
+  }
+
+  // Test that sharded and individual inputs matches what we wrote.
+  {
+    InputAndFeaturesVec elements;
+    ASSERT_OK(SampleSeedCorpusElementsFromSource(  //
+        SeedCorpusSource{
+            .dir_glob = std::string(kRelDir),
+            .num_recent_dirs = 1,
+            .shard_rel_glob = absl::StrCat("distilled-", kCovBin, ".*"),
+            // Intentionally try to match the shard files and test if they will
+            // be read as individual inputs.
+            .individual_input_rel_glob = "*",
+            .sampled_fraction_or_count = 1.0f,
+        },
+        kCovBin, kCovHash, elements));
+    EXPECT_EQ(elements.size(), 5);  // Non-empty inputs
+    EXPECT_THAT(elements, IsSupersetOf(kShardedInputs));
+    EXPECT_THAT(elements, IsSupersetOf(InputAndFeaturesVec{
+                              {{0, 1, 2}, {}},
+                              {{0, 1, 2, 3}, {}},
+                          }));
+  }
+}
+
 }  // namespace
 }  // namespace centipede