diff --git a/.cargo_vcs_info.json b/.cargo_vcs_info.json
index 600bdb7..4d6cfd1 100644
--- a/.cargo_vcs_info.json
+++ b/.cargo_vcs_info.json
@@ -1,5 +1,5 @@
 {
   "git": {
-    "sha1": "4b33ad40af65972c122a16180f51278830dff0a5"
+    "sha1": "a89115ac1105fa0c7c7d9cb5cdb479af36031aff"
   }
 }
diff --git a/Android.bp b/Android.bp
index 65d3b7a..fbfcb90 100644
--- a/Android.bp
+++ b/Android.bp
@@ -1,4 +1,4 @@
-// This file is generated by cargo2android.py --run --dependencies --device --features arbitrary-derive --patch patches/Android.bp.patch.
+// This file is generated by cargo2android.py --config cargo2android.json.
 // Do not modify this file as changes will be overridden on upgrade.
 
 package {
@@ -38,19 +38,13 @@
     name: "liblibfuzzer_sys",
     host_supported: true,
     crate_name: "libfuzzer_sys",
+    cargo_env_compat: true,
+    cargo_pkg_version: "0.4.2",
     srcs: ["src/lib.rs"],
     edition: "2018",
     features: ["arbitrary-derive"],
-    rlibs: [
+    rustlibs: [
         "libarbitrary",
+        "libonce_cell",
     ],
 }
-
-// dependent_library ["feature_list"]
-//   arbitrary-1.0.0 "derive,derive_arbitrary"
-//   cc-1.0.67
-//   derive_arbitrary-1.0.0
-//   proc-macro2-1.0.26 "default,proc-macro"
-//   quote-1.0.9 "default,proc-macro"
-//   syn-1.0.68 "clone-impls,default,derive,parsing,printing,proc-macro,quote"
-//   unicode-xid-0.2.1 "default"
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 26345b2..77be1fd 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -28,6 +28,46 @@
 
 --------------------------------------------------------------------------------
 
+## 0.4.2
+
+Released 2020-05-26.
+
+### Changed
+
+* Improved performance of checking for whether `cargo fuzz` is requesting the
+  `std::fmt::Debug` output of an input or not. This is always false during
+  regular fuzzing, so making this check faster should give slightly better
+  fuzzing throughput.
+
+--------------------------------------------------------------------------------
+
+## 0.4.1
+
+Released 2020-05-13.
+
+### Added
+
+* Added support for defining custom mutators. See [the documentation for the
+  `fuzz_mutator!`
+  macro](https://docs.rs/libfuzzer-sys/0.4.1/libfuzzer_sys/macro.fuzz_mutator.html)
+  for details.
+
+### Changed
+
+* Upgraded libfuzzer to llvm/llvm-project's 70cbc6d.
+
+--------------------------------------------------------------------------------
+
+## 0.4.0
+
+Released 2021-02-24.
+
+### Changed
+
+* The public `arbitrary` dependency was updated to version 1.0.
+
+--------------------------------------------------------------------------------
+
 ## 0.3.5
 
 Released 2020-11-18.
diff --git a/Cargo.toml b/Cargo.toml
index a1635d7..4be2136 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -13,7 +13,7 @@
 [package]
 edition = "2018"
 name = "libfuzzer-sys"
-version = "0.4.0"
+version = "0.4.2"
 authors = ["The rust-fuzz Project Developers"]
 description = "A wrapper around LLVM's libFuzzer runtime."
 readme = "./README.md"
@@ -21,6 +21,14 @@
 repository = "https://github.com/rust-fuzz/libfuzzer"
 [dependencies.arbitrary]
 version = "1"
+
+[dependencies.once_cell]
+version = "1"
+[dev-dependencies.flate2]
+version = "1.0.20"
+
+[dev-dependencies.rand]
+version = "0.8.3"
 [build-dependencies.cc]
 version = "1.0"
 
diff --git a/Cargo.toml.orig b/Cargo.toml.orig
index db1d7b3..72285b9 100644
--- a/Cargo.toml.orig
+++ b/Cargo.toml.orig
@@ -6,13 +6,25 @@
 name = "libfuzzer-sys"
 readme = "./README.md"
 repository = "https://github.com/rust-fuzz/libfuzzer"
-version = "0.4.0"
+version = "0.4.2"
 
 [dependencies]
 arbitrary = "1"
+once_cell = "1"
 
 [build-dependencies]
 cc = "1.0"
 
 [features]
 arbitrary-derive = ["arbitrary/derive"]
+
+[workspace]
+members = [
+  "./example",
+  "./example_arbitrary",
+  "./example_mutator",
+]
+
+[dev-dependencies]
+flate2 = "1.0.20"
+rand = "0.8.3"
diff --git a/LICENSE-APACHE b/LICENSE-APACHE
new file mode 100644
index 0000000..16fe87b
--- /dev/null
+++ b/LICENSE-APACHE
@@ -0,0 +1,201 @@
+                              Apache License
+                        Version 2.0, January 2004
+                     http://www.apache.org/licenses/
+
+TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+1. Definitions.
+
+   "License" shall mean the terms and conditions for use, reproduction,
+   and distribution as defined by Sections 1 through 9 of this document.
+
+   "Licensor" shall mean the copyright owner or entity authorized by
+   the copyright owner that is granting the License.
+
+   "Legal Entity" shall mean the union of the acting entity and all
+   other entities that control, are controlled by, or are under common
+   control with that entity. For the purposes of this definition,
+   "control" means (i) the power, direct or indirect, to cause the
+   direction or management of such entity, whether by contract or
+   otherwise, or (ii) ownership of fifty percent (50%) or more of the
+   outstanding shares, or (iii) beneficial ownership of such entity.
+
+   "You" (or "Your") shall mean an individual or Legal Entity
+   exercising permissions granted by this License.
+
+   "Source" form shall mean the preferred form for making modifications,
+   including but not limited to software source code, documentation
+   source, and configuration files.
+
+   "Object" form shall mean any form resulting from mechanical
+   transformation or translation of a Source form, including but
+   not limited to compiled object code, generated documentation,
+   and conversions to other media types.
+
+   "Work" shall mean the work of authorship, whether in Source or
+   Object form, made available under the License, as indicated by a
+   copyright notice that is included in or attached to the work
+   (an example is provided in the Appendix below).
+
+   "Derivative Works" shall mean any work, whether in Source or Object
+   form, that is based on (or derived from) the Work and for which the
+   editorial revisions, annotations, elaborations, or other modifications
+   represent, as a whole, an original work of authorship. For the purposes
+   of this License, Derivative Works shall not include works that remain
+   separable from, or merely link (or bind by name) to the interfaces of,
+   the Work and Derivative Works thereof.
+
+   "Contribution" shall mean any work of authorship, including
+   the original version of the Work and any modifications or additions
+   to that Work or Derivative Works thereof, that is intentionally
+   submitted to Licensor for inclusion in the Work by the copyright owner
+   or by an individual or Legal Entity authorized to submit on behalf of
+   the copyright owner. For the purposes of this definition, "submitted"
+   means any form of electronic, verbal, or written communication sent
+   to the Licensor or its representatives, including but not limited to
+   communication on electronic mailing lists, source code control systems,
+   and issue tracking systems that are managed by, or on behalf of, the
+   Licensor for the purpose of discussing and improving the Work, but
+   excluding communication that is conspicuously marked or otherwise
+   designated in writing by the copyright owner as "Not a Contribution."
+
+   "Contributor" shall mean Licensor and any individual or Legal Entity
+   on behalf of whom a Contribution has been received by Licensor and
+   subsequently incorporated within the Work.
+
+2. Grant of Copyright License. Subject to the terms and conditions of
+   this License, each Contributor hereby grants to You a perpetual,
+   worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+   copyright license to reproduce, prepare Derivative Works of,
+   publicly display, publicly perform, sublicense, and distribute the
+   Work and such Derivative Works in Source or Object form.
+
+3. Grant of Patent License. Subject to the terms and conditions of
+   this License, each Contributor hereby grants to You a perpetual,
+   worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+   (except as stated in this section) patent license to make, have made,
+   use, offer to sell, sell, import, and otherwise transfer the Work,
+   where such license applies only to those patent claims licensable
+   by such Contributor that are necessarily infringed by their
+   Contribution(s) alone or by combination of their Contribution(s)
+   with the Work to which such Contribution(s) was submitted. If You
+   institute patent litigation against any entity (including a
+   cross-claim or counterclaim in a lawsuit) alleging that the Work
+   or a Contribution incorporated within the Work constitutes direct
+   or contributory patent infringement, then any patent licenses
+   granted to You under this License for that Work shall terminate
+   as of the date such litigation is filed.
+
+4. Redistribution. You may reproduce and distribute copies of the
+   Work or Derivative Works thereof in any medium, with or without
+   modifications, and in Source or Object form, provided that You
+   meet the following conditions:
+
+   (a) You must give any other recipients of the Work or
+       Derivative Works a copy of this License; and
+
+   (b) You must cause any modified files to carry prominent notices
+       stating that You changed the files; and
+
+   (c) You must retain, in the Source form of any Derivative Works
+       that You distribute, all copyright, patent, trademark, and
+       attribution notices from the Source form of the Work,
+       excluding those notices that do not pertain to any part of
+       the Derivative Works; and
+
+   (d) If the Work includes a "NOTICE" text file as part of its
+       distribution, then any Derivative Works that You distribute must
+       include a readable copy of the attribution notices contained
+       within such NOTICE file, excluding those notices that do not
+       pertain to any part of the Derivative Works, in at least one
+       of the following places: within a NOTICE text file distributed
+       as part of the Derivative Works; within the Source form or
+       documentation, if provided along with the Derivative Works; or,
+       within a display generated by the Derivative Works, if and
+       wherever such third-party notices normally appear. The contents
+       of the NOTICE file are for informational purposes only and
+       do not modify the License. You may add Your own attribution
+       notices within Derivative Works that You distribute, alongside
+       or as an addendum to the NOTICE text from the Work, provided
+       that such additional attribution notices cannot be construed
+       as modifying the License.
+
+   You may add Your own copyright statement to Your modifications and
+   may provide additional or different license terms and conditions
+   for use, reproduction, or distribution of Your modifications, or
+   for any such Derivative Works as a whole, provided Your use,
+   reproduction, and distribution of the Work otherwise complies with
+   the conditions stated in this License.
+
+5. Submission of Contributions. Unless You explicitly state otherwise,
+   any Contribution intentionally submitted for inclusion in the Work
+   by You to the Licensor shall be under the terms and conditions of
+   this License, without any additional terms or conditions.
+   Notwithstanding the above, nothing herein shall supersede or modify
+   the terms of any separate license agreement you may have executed
+   with Licensor regarding such Contributions.
+
+6. Trademarks. This License does not grant permission to use the trade
+   names, trademarks, service marks, or product names of the Licensor,
+   except as required for reasonable and customary use in describing the
+   origin of the Work and reproducing the content of the NOTICE file.
+
+7. Disclaimer of Warranty. Unless required by applicable law or
+   agreed to in writing, Licensor provides the Work (and each
+   Contributor provides its Contributions) on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+   implied, including, without limitation, any warranties or conditions
+   of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+   PARTICULAR PURPOSE. You are solely responsible for determining the
+   appropriateness of using or redistributing the Work and assume any
+   risks associated with Your exercise of permissions under this License.
+
+8. Limitation of Liability. In no event and under no legal theory,
+   whether in tort (including negligence), contract, or otherwise,
+   unless required by applicable law (such as deliberate and grossly
+   negligent acts) or agreed to in writing, shall any Contributor be
+   liable to You for damages, including any direct, indirect, special,
+   incidental, or consequential damages of any character arising as a
+   result of this License or out of the use or inability to use the
+   Work (including but not limited to damages for loss of goodwill,
+   work stoppage, computer failure or malfunction, or any and all
+   other commercial damages or losses), even if such Contributor
+   has been advised of the possibility of such damages.
+
+9. Accepting Warranty or Additional Liability. While redistributing
+   the Work or Derivative Works thereof, You may choose to offer,
+   and charge a fee for, acceptance of support, warranty, indemnity,
+   or other liability obligations and/or rights consistent with this
+   License. However, in accepting such obligations, You may act only
+   on Your own behalf and on Your sole responsibility, not on behalf
+   of any other Contributor, and only if You agree to indemnify,
+   defend, and hold each Contributor harmless for any liability
+   incurred by, or claims asserted against, such Contributor by reason
+   of your accepting any such warranty or additional liability.
+
+END OF TERMS AND CONDITIONS
+
+APPENDIX: How to apply the Apache License to your work.
+
+   To apply the Apache License to your work, attach the following
+   boilerplate notice, with the fields enclosed by brackets "[]"
+   replaced with your own identifying information. (Don't include
+   the brackets!)  The text should be enclosed in the appropriate
+   comment syntax for the file format. We also recommend that a
+   file or class name and description of purpose be included on the
+   same "printed page" as the copyright notice for easier
+   identification within third-party archives.
+
+Copyright [yyyy] [name of copyright owner]
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+	http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
diff --git a/LICENSE-MIT b/LICENSE-MIT
new file mode 100644
index 0000000..25597d5
--- /dev/null
+++ b/LICENSE-MIT
@@ -0,0 +1,25 @@
+Copyright (c) 2010 The Rust Project Developers
+
+Permission is hereby granted, free of charge, to any
+person obtaining a copy of this software and associated
+documentation files (the "Software"), to deal in the
+Software without restriction, including without
+limitation the rights to use, copy, modify, merge,
+publish, distribute, sublicense, and/or sell copies of
+the Software, and to permit persons to whom the Software
+is furnished to do so, subject to the following
+conditions:
+
+The above copyright notice and this permission notice
+shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF
+ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED
+TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
+PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT
+SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
+CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
+OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
+IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+DEALINGS IN THE SOFTWARE.
diff --git a/METADATA b/METADATA
index bcae820..abca0eb 100644
--- a/METADATA
+++ b/METADATA
@@ -7,13 +7,13 @@
   }
   url {
     type: ARCHIVE
-    value: "https://static.crates.io/crates/libfuzzer-sys/libfuzzer-sys-0.4.0.crate"
+    value: "https://static.crates.io/crates/libfuzzer-sys/libfuzzer-sys-0.4.2.crate"
   }
-  version: "0.4.0"
+  version: "0.4.2"
   license_type: NOTICE
   last_upgrade_date {
     year: 2021
-    month: 4
-    day: 1
+    month: 6
+    day: 21
   }
 }
diff --git a/README.md b/README.md
index 1d01871..8bdba6e 100644
--- a/README.md
+++ b/README.md
@@ -29,7 +29,7 @@
 
 ```toml
 [dependencies]
-libfuzzer-sys = "0.3.0"
+libfuzzer-sys = "0.4.0"
 your_crate = { path = "../path/to/your/crate" }
 ```
 
diff --git a/cargo2android.json b/cargo2android.json
new file mode 100644
index 0000000..01e723e
--- /dev/null
+++ b/cargo2android.json
@@ -0,0 +1,11 @@
+{
+  "dependencies": true,
+  "device": true,
+  "features": "arbitrary-derive",
+  "force-rlib": true,
+  "lib-blocklist": [
+    "fuzzer",
+    "stdc++"
+  ],
+  "run": true
+}
\ No newline at end of file
diff --git a/ci/script.sh b/ci/script.sh
index 25914fd..91fad80 100755
--- a/ci/script.sh
+++ b/ci/script.sh
@@ -5,6 +5,8 @@
 
 export CARGO_TARGET_DIR=$(pwd)/target
 
+cargo test --doc
+
 pushd ./example
 cargo rustc \
       --release \
@@ -39,3 +41,20 @@
 cat $(pwd)/debug_output
 grep -q Rgb $(pwd)/debug_output
 popd
+
+pushd ./example_mutator
+cargo rustc \
+      --release \
+      -- \
+      -Cpasses='sancov' \
+      -Cllvm-args=-sanitizer-coverage-level=3 \
+      -Cllvm-args=-sanitizer-coverage-trace-compares \
+      -Cllvm-args=-sanitizer-coverage-inline-8bit-counters \
+      -Cllvm-args=-sanitizer-coverage-stack-depth \
+      -Cllvm-args=-sanitizer-coverage-trace-geps \
+      -Cllvm-args=-sanitizer-coverage-prune-blocks=0 \
+      -Zsanitizer=address
+(! $CARGO_TARGET_DIR/release/example_mutator -runs=10000000)
+popd
+
+echo "All good!"
diff --git a/libfuzzer/FuzzerBuiltins.h b/libfuzzer/FuzzerBuiltins.h
index 4c0ada8..ce0bd5c 100644
--- a/libfuzzer/FuzzerBuiltins.h
+++ b/libfuzzer/FuzzerBuiltins.h
@@ -26,7 +26,6 @@
 inline uint64_t Bswap(uint64_t x) { return __builtin_bswap64(x); }
 
 inline uint32_t Clzll(unsigned long long X) { return __builtin_clzll(X); }
-inline uint32_t Clz(unsigned long long X) { return __builtin_clz(X); }
 inline int Popcountll(unsigned long long X) { return __builtin_popcountll(X); }
 
 }  // namespace fuzzer
diff --git a/libfuzzer/FuzzerBuiltinsMsvc.h b/libfuzzer/FuzzerBuiltinsMsvc.h
index c5bec97..ab191b6 100644
--- a/libfuzzer/FuzzerBuiltinsMsvc.h
+++ b/libfuzzer/FuzzerBuiltinsMsvc.h
@@ -52,12 +52,6 @@
   return 64;
 }
 
-inline uint32_t Clz(uint32_t X) {
-  unsigned long LeadZeroIdx = 0;
-  if (_BitScanReverse(&LeadZeroIdx, X)) return 31 - LeadZeroIdx;
-  return 32;
-}
-
 inline int Popcountll(unsigned long long X) {
 #if !defined(_M_ARM) && !defined(_M_X64)
   return __popcnt(X) + __popcnt(X >> 32);
diff --git a/libfuzzer/FuzzerCorpus.h b/libfuzzer/FuzzerCorpus.h
index daea4f5..f8c1260 100644
--- a/libfuzzer/FuzzerCorpus.h
+++ b/libfuzzer/FuzzerCorpus.h
@@ -44,7 +44,7 @@
   // Power schedule.
   bool NeedsEnergyUpdate = false;
   double Energy = 0.0;
-  size_t SumIncidence = 0;
+  double SumIncidence = 0.0;
   Vector<std::pair<uint32_t, uint16_t>> FeatureFreqs;
 
   // Delete feature Idx and its frequency from FeatureFreqs.
@@ -74,27 +74,28 @@
   void UpdateEnergy(size_t GlobalNumberOfFeatures, bool ScalePerExecTime,
                     std::chrono::microseconds AverageUnitExecutionTime) {
     Energy = 0.0;
-    SumIncidence = 0;
+    SumIncidence = 0.0;
 
     // Apply add-one smoothing to locally discovered features.
     for (auto F : FeatureFreqs) {
-      size_t LocalIncidence = F.second + 1;
-      Energy -= LocalIncidence * logl(LocalIncidence);
+      double LocalIncidence = F.second + 1;
+      Energy -= LocalIncidence * log(LocalIncidence);
       SumIncidence += LocalIncidence;
     }
 
     // Apply add-one smoothing to locally undiscovered features.
-    //   PreciseEnergy -= 0; // since logl(1.0) == 0)
-    SumIncidence += (GlobalNumberOfFeatures - FeatureFreqs.size());
+    //   PreciseEnergy -= 0; // since log(1.0) == 0)
+    SumIncidence +=
+        static_cast<double>(GlobalNumberOfFeatures - FeatureFreqs.size());
 
     // Add a single locally abundant feature apply add-one smoothing.
-    size_t AbdIncidence = NumExecutedMutations + 1;
-    Energy -= AbdIncidence * logl(AbdIncidence);
+    double AbdIncidence = static_cast<double>(NumExecutedMutations + 1);
+    Energy -= AbdIncidence * log(AbdIncidence);
     SumIncidence += AbdIncidence;
 
     // Normalize.
     if (SumIncidence != 0)
-      Energy = (Energy / SumIncidence) + logl(SumIncidence);
+      Energy = Energy / SumIncidence + log(SumIncidence);
 
     if (ScalePerExecTime) {
       // Scaling to favor inputs with lower execution time.
@@ -213,6 +214,8 @@
     assert(!U.empty());
     if (FeatureDebug)
       Printf("ADD_TO_CORPUS %zd NF %zd\n", Inputs.size(), NumFeatures);
+    // Inputs.size() is cast to uint32_t below.
+    assert(Inputs.size() < std::numeric_limits<uint32_t>::max());
     Inputs.push_back(new InputInfo());
     InputInfo &II = *Inputs.back();
     II.U = U;
@@ -224,7 +227,7 @@
     II.HasFocusFunction = HasFocusFunction;
     // Assign maximal energy to the new seed.
     II.Energy = RareFeatures.empty() ? 1.0 : log(RareFeatures.size());
-    II.SumIncidence = RareFeatures.size();
+    II.SumIncidence = static_cast<double>(RareFeatures.size());
     II.NeedsEnergyUpdate = false;
     std::sort(II.UniqFeatureSet.begin(), II.UniqFeatureSet.end());
     ComputeSHA1(U.data(), U.size(), II.Sha1);
@@ -399,7 +402,7 @@
       // Zero energy seeds will never be fuzzed and remain zero energy.
       if (II->Energy > 0.0) {
         II->SumIncidence += 1;
-        II->Energy += logl(II->SumIncidence) / II->SumIncidence;
+        II->Energy += log(II->SumIncidence) / II->SumIncidence;
       }
     }
 
@@ -426,7 +429,8 @@
       NumUpdatedFeatures++;
       if (FeatureDebug)
         Printf("ADD FEATURE %zd sz %d\n", Idx, NewSize);
-      SmallestElementPerFeature[Idx] = Inputs.size();
+      // Inputs.size() is guaranteed to be less than UINT32_MAX by AddToCorpus.
+      SmallestElementPerFeature[Idx] = static_cast<uint32_t>(Inputs.size());
       InputSizesPerFeature[Idx] = NewSize;
       return true;
     }
@@ -464,7 +468,7 @@
 
   static const bool FeatureDebug = false;
 
-  size_t GetFeature(size_t Idx) const { return InputSizesPerFeature[Idx]; }
+  uint32_t GetFeature(size_t Idx) const { return InputSizesPerFeature[Idx]; }
 
   void ValidateFeatureSet() {
     if (FeatureDebug)
@@ -539,9 +543,11 @@
 
     if (VanillaSchedule) {
       for (size_t i = 0; i < N; i++)
-        Weights[i] = Inputs[i]->NumFeatures
-                         ? (i + 1) * (Inputs[i]->HasFocusFunction ? 1000 : 1)
-                         : 0.;
+        Weights[i] =
+            Inputs[i]->NumFeatures
+                ? static_cast<double>((i + 1) *
+                                      (Inputs[i]->HasFocusFunction ? 1000 : 1))
+                : 0.;
     }
 
     if (FeatureDebug) {
diff --git a/libfuzzer/FuzzerDataFlowTrace.cpp b/libfuzzer/FuzzerDataFlowTrace.cpp
index 0e9cdf7..23d4225 100644
--- a/libfuzzer/FuzzerDataFlowTrace.cpp
+++ b/libfuzzer/FuzzerDataFlowTrace.cpp
@@ -60,6 +60,7 @@
       CoveredBlocks.push_back(BB);
     }
     if (CoveredBlocks.empty()) return false;
+    // Ensures no CoverageVector is longer than UINT32_MAX.
     uint32_t NumBlocks = CoveredBlocks.back();
     CoveredBlocks.pop_back();
     for (auto BB : CoveredBlocks)
@@ -200,7 +201,8 @@
     Printf("INFO: AUTOFOCUS: %zd %s\n", FocusFuncIdx,
            FunctionNames[FocusFuncIdx].c_str());
     for (size_t i = 0; i < NumFunctions; i++) {
-      if (!Weights[i]) continue;
+      if (Weights[i] == 0.0)
+        continue;
       Printf("  [%zd] W %g\tBB-tot %u\tBB-cov %u\tEntryFreq %u:\t%s\n", i,
              Weights[i], Coverage.GetNumberOfBlocks(i),
              Coverage.GetNumberOfCoveredBlocks(i), Coverage.GetCounter(i, 0),
diff --git a/libfuzzer/FuzzerDataFlowTrace.h b/libfuzzer/FuzzerDataFlowTrace.h
index d6e3de3..07c03bb 100644
--- a/libfuzzer/FuzzerDataFlowTrace.h
+++ b/libfuzzer/FuzzerDataFlowTrace.h
@@ -42,7 +42,8 @@
                     const Vector<SizedFile> &CorporaFiles);
 
 class BlockCoverage {
- public:
+public:
+  // These functions guarantee no CoverageVector is longer than UINT32_MAX.
   bool AppendCoverage(std::istream &IN);
   bool AppendCoverage(const std::string &S);
 
@@ -50,7 +51,8 @@
 
   uint32_t GetCounter(size_t FunctionId, size_t BasicBlockId) {
     auto It = Functions.find(FunctionId);
-    if (It == Functions.end()) return 0;
+    if (It == Functions.end())
+      return 0;
     const auto &Counters = It->second;
     if (BasicBlockId < Counters.size())
       return Counters[BasicBlockId];
@@ -61,7 +63,7 @@
     auto It = Functions.find(FunctionId);
     if (It == Functions.end()) return 0;
     const auto &Counters = It->second;
-    return Counters.size();
+    return static_cast<uint32_t>(Counters.size());
   }
 
   uint32_t GetNumberOfCoveredBlocks(size_t FunctionId) {
@@ -78,8 +80,7 @@
   Vector<double> FunctionWeights(size_t NumFunctions) const;
   void clear() { Functions.clear(); }
 
- private:
-
+private:
   typedef Vector<uint32_t> CoverageVector;
 
   uint32_t NumberOfCoveredBlocks(const CoverageVector &Counters) const {
@@ -91,7 +92,8 @@
   }
 
   uint32_t NumberOfUncoveredBlocks(const CoverageVector &Counters) const {
-    return Counters.size() - NumberOfCoveredBlocks(Counters);
+    return static_cast<uint32_t>(Counters.size()) -
+           NumberOfCoveredBlocks(Counters);
   }
 
   uint32_t SmallestNonZeroCounter(const CoverageVector &Counters) const {
diff --git a/libfuzzer/FuzzerDictionary.h b/libfuzzer/FuzzerDictionary.h
index 301c5d9..db55907 100644
--- a/libfuzzer/FuzzerDictionary.h
+++ b/libfuzzer/FuzzerDictionary.h
@@ -23,12 +23,14 @@
 public:
   static const size_t kMaxSize = kMaxSizeT;
   FixedWord() {}
-  FixedWord(const uint8_t *B, uint8_t S) { Set(B, S); }
+  FixedWord(const uint8_t *B, size_t S) { Set(B, S); }
 
-  void Set(const uint8_t *B, uint8_t S) {
+  void Set(const uint8_t *B, size_t S) {
+    static_assert(kMaxSizeT <= std::numeric_limits<uint8_t>::max(),
+                  "FixedWord::kMaxSizeT cannot fit in a uint8_t.");
     assert(S <= kMaxSize);
     memcpy(Data, B, S);
-    Size = S;
+    Size = static_cast<uint8_t>(S);
   }
 
   bool operator==(const FixedWord<kMaxSize> &w) const {
diff --git a/libfuzzer/FuzzerDriver.cpp b/libfuzzer/FuzzerDriver.cpp
index 447cafc..ceaa907 100644
--- a/libfuzzer/FuzzerDriver.cpp
+++ b/libfuzzer/FuzzerDriver.cpp
@@ -159,14 +159,14 @@
     const char *Str = FlagValue(Param, Name);
     if (Str)  {
       if (FlagDescriptions[F].IntFlag) {
-        int Val = MyStol(Str);
-        *FlagDescriptions[F].IntFlag = Val;
+        auto Val = MyStol(Str);
+        *FlagDescriptions[F].IntFlag = static_cast<int>(Val);
         if (Flags.verbosity >= 2)
           Printf("Flag: %s %d\n", Name, Val);
         return true;
       } else if (FlagDescriptions[F].UIntFlag) {
-        unsigned int Val = std::stoul(Str);
-        *FlagDescriptions[F].UIntFlag = Val;
+        auto Val = std::stoul(Str);
+        *FlagDescriptions[F].UIntFlag = static_cast<unsigned int>(Val);
         if (Flags.verbosity >= 2)
           Printf("Flag: %s %u\n", Name, Val);
         return true;
@@ -789,8 +789,8 @@
   unsigned Seed = Flags.seed;
   // Initialize Seed.
   if (Seed == 0)
-    Seed =
-        std::chrono::system_clock::now().time_since_epoch().count() + GetPid();
+    Seed = static_cast<unsigned>(
+        std::chrono::system_clock::now().time_since_epoch().count() + GetPid());
   if (Flags.verbosity)
     Printf("INFO: Seed: %u\n", Seed);
 
diff --git a/libfuzzer/FuzzerFork.cpp b/libfuzzer/FuzzerFork.cpp
index 84725d2..5134a5d 100644
--- a/libfuzzer/FuzzerFork.cpp
+++ b/libfuzzer/FuzzerFork.cpp
@@ -142,7 +142,9 @@
         CollectDFT(SF);
       }
       auto Time2 = std::chrono::system_clock::now();
-      Job->DftTimeInSeconds = duration_cast<seconds>(Time2 - Time1).count();
+      auto DftTimeInSeconds = duration_cast<seconds>(Time2 - Time1).count();
+      assert(DftTimeInSeconds < std::numeric_limits<int>::max());
+      Job->DftTimeInSeconds = static_cast<int>(DftTimeInSeconds);
     }
     if (!Seeds.empty()) {
       Job->SeedListPath =
@@ -314,8 +316,11 @@
       Env.Files.push_back(File.File);
   } else {
     auto CFPath = DirPlusFile(Env.TempDir, "merge.txt");
-    CrashResistantMerge(Env.Args, {}, SeedFiles, &Env.Files, {}, &Env.Features,
-                        {}, &Env.Cov, CFPath, false);
+    Set<uint32_t> NewFeatures, NewCov;
+    CrashResistantMerge(Env.Args, {}, SeedFiles, &Env.Files, Env.Features,
+                        &NewFeatures, Env.Cov, &NewCov, CFPath, false);
+    Env.Features.insert(NewFeatures.begin(), NewFeatures.end());
+    Env.Cov.insert(NewFeatures.begin(), NewFeatures.end());
     RemoveFile(CFPath);
   }
   Printf("INFO: -fork=%d: %zd seed inputs, starting to fuzz in %s\n", NumJobs,
diff --git a/libfuzzer/FuzzerIO.cpp b/libfuzzer/FuzzerIO.cpp
index 54a7219..7f149ac 100644
--- a/libfuzzer/FuzzerIO.cpp
+++ b/libfuzzer/FuzzerIO.cpp
@@ -90,8 +90,9 @@
   fclose(Out);
 }
 
-void ReadDirToVectorOfUnits(const char *Path, Vector<Unit> *V,
-                            long *Epoch, size_t MaxSize, bool ExitOnError) {
+void ReadDirToVectorOfUnits(const char *Path, Vector<Unit> *V, long *Epoch,
+                            size_t MaxSize, bool ExitOnError,
+                            Vector<std::string> *VPaths) {
   long E = Epoch ? *Epoch : 0;
   Vector<std::string> Files;
   ListFilesInDirRecursive(Path, Epoch, &Files, /*TopDir*/true);
@@ -103,12 +104,14 @@
     if ((NumLoaded & (NumLoaded - 1)) == 0 && NumLoaded >= 1024)
       Printf("Loaded %zd/%zd files from %s\n", NumLoaded, Files.size(), Path);
     auto S = FileToVector(X, MaxSize, ExitOnError);
-    if (!S.empty())
+    if (!S.empty()) {
       V->push_back(S);
+      if (VPaths)
+        VPaths->push_back(X);
+    }
   }
 }
 
-
 void GetSizedFilesFromDir(const std::string &Dir, Vector<SizedFile> *V) {
   Vector<std::string> Files;
   ListFilesInDirRecursive(Dir, 0, &Files, /*TopDir*/true);
diff --git a/libfuzzer/FuzzerIO.h b/libfuzzer/FuzzerIO.h
index abd2511..bde1826 100644
--- a/libfuzzer/FuzzerIO.h
+++ b/libfuzzer/FuzzerIO.h
@@ -32,8 +32,9 @@
 void AppendToFile(const uint8_t *Data, size_t Size, const std::string &Path);
 void AppendToFile(const std::string &Data, const std::string &Path);
 
-void ReadDirToVectorOfUnits(const char *Path, Vector<Unit> *V,
-                            long *Epoch, size_t MaxSize, bool ExitOnError);
+void ReadDirToVectorOfUnits(const char *Path, Vector<Unit> *V, long *Epoch,
+                            size_t MaxSize, bool ExitOnError,
+                            Vector<std::string> *VPaths = 0);
 
 // Returns "Dir/FileName" or equivalent for the current OS.
 std::string DirPlusFile(const std::string &DirPath,
diff --git a/libfuzzer/FuzzerLoop.cpp b/libfuzzer/FuzzerLoop.cpp
index 6e3bf44..86a78ab 100644
--- a/libfuzzer/FuzzerLoop.cpp
+++ b/libfuzzer/FuzzerLoop.cpp
@@ -414,19 +414,25 @@
   if (Options.OutputCorpus.empty() || !Options.ReloadIntervalSec)
     return;
   Vector<Unit> AdditionalCorpus;
-  ReadDirToVectorOfUnits(Options.OutputCorpus.c_str(), &AdditionalCorpus,
-                         &EpochOfLastReadOfOutputCorpus, MaxSize,
-                         /*ExitOnError*/ false);
+  Vector<std::string> AdditionalCorpusPaths;
+  ReadDirToVectorOfUnits(
+      Options.OutputCorpus.c_str(), &AdditionalCorpus,
+      &EpochOfLastReadOfOutputCorpus, MaxSize,
+      /*ExitOnError*/ false,
+      (Options.Verbosity >= 2 ? &AdditionalCorpusPaths : nullptr));
   if (Options.Verbosity >= 2)
     Printf("Reload: read %zd new units.\n", AdditionalCorpus.size());
   bool Reloaded = false;
-  for (auto &U : AdditionalCorpus) {
+  for (size_t i = 0; i != AdditionalCorpus.size(); ++i) {
+    auto &U = AdditionalCorpus[i];
     if (U.size() > MaxSize)
       U.resize(MaxSize);
     if (!Corpus.HasUnit(U)) {
       if (RunOne(U.data(), U.size())) {
         CheckExitOnSrcPosOrItem();
         Reloaded = true;
+        if (Options.Verbosity >= 2)
+          Printf("Reloaded %s\n", AdditionalCorpusPaths[i].c_str());
       }
     }
   }
@@ -440,8 +446,9 @@
   if (!(TotalNumberOfRuns & (TotalNumberOfRuns - 1)) &&
       secondsSinceProcessStartUp() >= 2)
     PrintStats("pulse ");
-  if (TimeOfUnit > TimeOfLongestUnitInSeconds * 1.1 &&
-      TimeOfUnit >= Options.ReportSlowUnits) {
+  auto Threshhold =
+      static_cast<long>(static_cast<double>(TimeOfLongestUnitInSeconds) * 1.1);
+  if (TimeOfUnit > Threshhold && TimeOfUnit >= Options.ReportSlowUnits) {
     TimeOfLongestUnitInSeconds = TimeOfUnit;
     Printf("Slowest unit: %zd s:\n", TimeOfLongestUnitInSeconds);
     WriteUnitToFileWithPrefix({Data, Data + Size}, "slow-unit-");
@@ -501,6 +508,8 @@
                     bool *FoundUniqFeatures) {
   if (!Size)
     return false;
+  // Largest input length should be INT_MAX.
+  assert(Size < std::numeric_limits<uint32_t>::max());
 
   ExecuteCallback(Data, Size);
   auto TimeOfUnit = duration_cast<microseconds>(UnitStopTime - UnitStartTime);
@@ -508,8 +517,8 @@
   UniqFeatureSetTmp.clear();
   size_t FoundUniqFeaturesOfII = 0;
   size_t NumUpdatesBefore = Corpus.NumFeatureUpdates();
-  TPC.CollectFeatures([&](size_t Feature) {
-    if (Corpus.AddFeature(Feature, Size, Options.Shrink))
+  TPC.CollectFeatures([&](uint32_t Feature) {
+    if (Corpus.AddFeature(Feature, static_cast<uint32_t>(Size), Options.Shrink))
       UniqFeatureSetTmp.push_back(Feature);
     if (Options.Entropic)
       Corpus.UpdateFeatureFrequency(II, Feature);
@@ -575,7 +584,10 @@
          !memcmp(A + Size - Limit / 2, B + Size - Limit / 2, Limit / 2);
 }
 
-void Fuzzer::ExecuteCallback(const uint8_t *Data, size_t Size) {
+// This method is not inlined because it would cause a test to fail where it
+// is part of the stack unwinding. See D97975 for details.
+ATTRIBUTE_NOINLINE void Fuzzer::ExecuteCallback(const uint8_t *Data,
+                                                size_t Size) {
   TPC.RecordInitialStack();
   TotalNumberOfRuns++;
   assert(InFuzzingThread());
diff --git a/libfuzzer/FuzzerMerge.cpp b/libfuzzer/FuzzerMerge.cpp
index e3ad8b3..162453c 100644
--- a/libfuzzer/FuzzerMerge.cpp
+++ b/libfuzzer/FuzzerMerge.cpp
@@ -82,9 +82,9 @@
   while (std::getline(IS, Line, '\n')) {
     std::istringstream ISS1(Line);
     std::string Marker;
-    size_t N;
-    ISS1 >> Marker;
-    ISS1 >> N;
+    uint32_t N;
+    if (!(ISS1 >> Marker) || !(ISS1 >> N))
+      return false;
     if (Marker == "STARTED") {
       // STARTED FILE_ID FILE_SIZE
       if (ExpectedStartMarker != N)
@@ -137,6 +137,8 @@
                      const Set<uint32_t> &InitialCov, Set<uint32_t> *NewCov,
                      Vector<std::string> *NewFiles) {
   NewFiles->clear();
+  NewFeatures->clear();
+  NewCov->clear();
   assert(NumFilesInFirstCorpus <= Files.size());
   Set<uint32_t> AllFeatures = InitialFeatures;
 
diff --git a/libfuzzer/FuzzerMutate.cpp b/libfuzzer/FuzzerMutate.cpp
index cf34a9f..4650f1b 100644
--- a/libfuzzer/FuzzerMutate.cpp
+++ b/libfuzzer/FuzzerMutate.cpp
@@ -61,14 +61,20 @@
 }
 
 static char RandCh(Random &Rand) {
-  if (Rand.RandBool()) return Rand(256);
+  if (Rand.RandBool())
+    return static_cast<char>(Rand(256));
   const char Special[] = "!*'();:@&=+$,/?%#[]012Az-`~.\xff\x00";
   return Special[Rand(sizeof(Special) - 1)];
 }
 
 size_t MutationDispatcher::Mutate_Custom(uint8_t *Data, size_t Size,
                                          size_t MaxSize) {
-  return EF->LLVMFuzzerCustomMutator(Data, Size, MaxSize, Rand.Rand());
+  if (EF->__msan_unpoison)
+    EF->__msan_unpoison(Data, Size);
+  if (EF->__msan_unpoison_param)
+    EF->__msan_unpoison_param(4);
+  return EF->LLVMFuzzerCustomMutator(Data, Size, MaxSize,
+                                     Rand.Rand<unsigned int>());
 }
 
 size_t MutationDispatcher::Mutate_CustomCrossOver(uint8_t *Data, size_t Size,
@@ -81,8 +87,18 @@
     return 0;
   CustomCrossOverInPlaceHere.resize(MaxSize);
   auto &U = CustomCrossOverInPlaceHere;
+
+  if (EF->__msan_unpoison) {
+    EF->__msan_unpoison(Data, Size);
+    EF->__msan_unpoison(Other.data(), Other.size());
+    EF->__msan_unpoison(U.data(), U.size());
+  }
+  if (EF->__msan_unpoison_param)
+    EF->__msan_unpoison_param(7);
   size_t NewSize = EF->LLVMFuzzerCustomCrossOver(
-      Data, Size, Other.data(), Other.size(), U.data(), U.size(), Rand.Rand());
+      Data, Size, Other.data(), Other.size(), U.data(), U.size(),
+      Rand.Rand<unsigned int>());
+
   if (!NewSize)
     return 0;
   assert(NewSize <= MaxSize && "CustomCrossOver returned overisized unit");
@@ -135,7 +151,8 @@
   // Insert new values at Data[Idx].
   memmove(Data + Idx + N, Data + Idx, Size - Idx);
   // Give preference to 0x00 and 0xff.
-  uint8_t Byte = Rand.RandBool() ? Rand(256) : (Rand.RandBool() ? 0 : 255);
+  uint8_t Byte = static_cast<uint8_t>(
+      Rand.RandBool() ? Rand(256) : (Rand.RandBool() ? 0 : 255));
   for (size_t i = 0; i < N; i++)
     Data[Idx + i] = Byte;
   return Size + N;
@@ -178,7 +195,8 @@
     Size += W.size();
   } else {  // Overwrite some bytes with W.
     if (W.size() > Size) return 0;
-    size_t Idx = UsePositionHint ? DE.GetPositionHint() : Rand(Size - W.size());
+    size_t Idx =
+        UsePositionHint ? DE.GetPositionHint() : Rand(Size + 1 - W.size());
     memcpy(Data + Idx, W.data(), W.size());
   }
   return Size;
@@ -227,8 +245,8 @@
     T Arg1, T Arg2, const uint8_t *Data, size_t Size) {
   if (Rand.RandBool()) Arg1 = Bswap(Arg1);
   if (Rand.RandBool()) Arg2 = Bswap(Arg2);
-  T Arg1Mutation = Arg1 + Rand(-1, 1);
-  T Arg2Mutation = Arg2 + Rand(-1, 1);
+  T Arg1Mutation = static_cast<T>(Arg1 + Rand(-1, 1));
+  T Arg2Mutation = static_cast<T>(Arg2 + Rand(-1, 1));
   return MakeDictionaryEntryFromCMP(&Arg1, &Arg2, &Arg1Mutation, &Arg2Mutation,
                                     sizeof(Arg1), Data, Size);
 }
@@ -245,23 +263,23 @@
   DictionaryEntry DE;
   switch (Rand(4)) {
   case 0: {
-    auto X = TPC.TORC8.Get(Rand.Rand());
+    auto X = TPC.TORC8.Get(Rand.Rand<size_t>());
     DE = MakeDictionaryEntryFromCMP(X.A, X.B, Data, Size);
   } break;
   case 1: {
-    auto X = TPC.TORC4.Get(Rand.Rand());
+    auto X = TPC.TORC4.Get(Rand.Rand<size_t>());
     if ((X.A >> 16) == 0 && (X.B >> 16) == 0 && Rand.RandBool())
       DE = MakeDictionaryEntryFromCMP((uint16_t)X.A, (uint16_t)X.B, Data, Size);
     else
       DE = MakeDictionaryEntryFromCMP(X.A, X.B, Data, Size);
   } break;
   case 2: {
-    auto X = TPC.TORCW.Get(Rand.Rand());
+    auto X = TPC.TORCW.Get(Rand.Rand<size_t>());
     DE = MakeDictionaryEntryFromCMP(X.A, X.B, Data, Size);
   } break;
   case 3: if (Options.UseMemmem) {
-    auto X = TPC.MMT.Get(Rand.Rand());
-    DE = DictionaryEntry(X);
+      auto X = TPC.MMT.Get(Rand.Rand<size_t>());
+      DE = DictionaryEntry(X);
   } break;
   default:
     assert(0);
@@ -387,12 +405,12 @@
   assert(Off + sizeof(T) <= Size);
   T Val;
   if (Off < 64 && !Rand(4)) {
-    Val = Size;
+    Val = static_cast<T>(Size);
     if (Rand.RandBool())
       Val = Bswap(Val);
   } else {
     memcpy(&Val, Data + Off, sizeof(Val));
-    T Add = Rand(21);
+    T Add = static_cast<T>(Rand(21));
     Add -= 10;
     if (Rand.RandBool())
       Val = Bswap(T(Bswap(Val) + Add)); // Add assuming different endiannes.
@@ -462,7 +480,7 @@
     assert(DE->GetW().size());
     // Linear search is fine here as this happens seldom.
     if (!PersistentAutoDictionary.ContainsWord(DE->GetW()))
-      PersistentAutoDictionary.push_back({DE->GetW(), 1});
+      PersistentAutoDictionary.push_back(*DE);
   }
 }
 
diff --git a/libfuzzer/FuzzerRandom.h b/libfuzzer/FuzzerRandom.h
index 659283e..ad6c07e 100644
--- a/libfuzzer/FuzzerRandom.h
+++ b/libfuzzer/FuzzerRandom.h
@@ -18,18 +18,27 @@
  public:
   Random(unsigned int seed) : std::minstd_rand(seed) {}
   result_type operator()() { return this->std::minstd_rand::operator()(); }
-  size_t Rand() { return this->operator()(); }
-  size_t RandBool() { return Rand() % 2; }
+  template <typename T>
+  typename std::enable_if<std::is_integral<T>::value, T>::type Rand() {
+    return static_cast<T>(this->operator()());
+  }
+  size_t RandBool() { return this->operator()() % 2; }
   size_t SkewTowardsLast(size_t n) {
     size_t T = this->operator()(n * n);
-    size_t Res = sqrt(T);
+    size_t Res = static_cast<size_t>(sqrt(T));
     return Res;
   }
-  size_t operator()(size_t n) { return n ? Rand() % n : 0; }
-  intptr_t operator()(intptr_t From, intptr_t To) {
+  template <typename T>
+  typename std::enable_if<std::is_integral<T>::value, T>::type operator()(T n) {
+    return n ? Rand<T>() % n : 0;
+  }
+  template <typename T>
+  typename std::enable_if<std::is_integral<T>::value, T>::type
+  operator()(T From, T To) {
     assert(From < To);
-    intptr_t RangeSize = To - From + 1;
-    return operator()(RangeSize) + From;
+    auto RangeSize = static_cast<unsigned long long>(To) -
+                     static_cast<unsigned long long>(From) + 1;
+    return static_cast<T>(this->operator()(RangeSize) + From);
   }
 };
 
diff --git a/libfuzzer/FuzzerSHA1.cpp b/libfuzzer/FuzzerSHA1.cpp
index 2005dc7..b05655c 100644
--- a/libfuzzer/FuzzerSHA1.cpp
+++ b/libfuzzer/FuzzerSHA1.cpp
@@ -134,12 +134,13 @@
 	s->state[4] += e;
 }
 
-void sha1_addUncounted(sha1nfo *s, uint8_t data) {
-	uint8_t * const b = (uint8_t*) s->buffer;
+// Adds the least significant byte of |data|.
+void sha1_addUncounted(sha1nfo *s, uint32_t data) {
+  uint8_t *const b = (uint8_t *)s->buffer;
 #ifdef SHA_BIG_ENDIAN
-	b[s->bufferOffset] = data;
+  b[s->bufferOffset] = static_cast<uint8_t>(data);
 #else
-	b[s->bufferOffset ^ 3] = data;
+  b[s->bufferOffset ^ 3] = static_cast<uint8_t>(data);
 #endif
 	s->bufferOffset++;
 	if (s->bufferOffset == BLOCK_LENGTH) {
diff --git a/libfuzzer/FuzzerTracePC.cpp b/libfuzzer/FuzzerTracePC.cpp
index 91e94d8..d808b9b 100644
--- a/libfuzzer/FuzzerTracePC.cpp
+++ b/libfuzzer/FuzzerTracePC.cpp
@@ -106,6 +106,15 @@
   }
   if (size_t NumExtraCounters = ExtraCountersEnd() - ExtraCountersBegin())
     Printf("INFO: %zd Extra Counters\n", NumExtraCounters);
+
+  size_t MaxFeatures = CollectFeatures([](uint32_t) {});
+  if (MaxFeatures > std::numeric_limits<uint32_t>::max())
+    Printf("WARNING: The coverage PC tables may produce up to %zu features.\n"
+           "This exceeds the maximum 32-bit value. Some features may be\n"
+           "ignored, and fuzzing may become less precise. If possible,\n"
+           "consider refactoring the fuzzer into several smaller fuzzers\n"
+           "linked against only a portion of the current target.\n",
+           MaxFeatures);
 }
 
 ATTRIBUTE_NO_SANITIZE_ALL
@@ -356,7 +365,7 @@
   uint8_t HammingDistance = 0;
   for (; I < Len; I++) {
     if (B1[I] != B2[I] || (StopAtZero && B1[I] == 0)) {
-      HammingDistance = Popcountll(B1[I] ^ B2[I]);
+      HammingDistance = static_cast<uint8_t>(Popcountll(B1[I] ^ B2[I]));
       break;
     }
   }
diff --git a/libfuzzer/FuzzerTracePC.h b/libfuzzer/FuzzerTracePC.h
index 0090923..a937329 100644
--- a/libfuzzer/FuzzerTracePC.h
+++ b/libfuzzer/FuzzerTracePC.h
@@ -54,7 +54,7 @@
   void Add(const uint8_t *Data, size_t Size) {
     if (Size <= 2) return;
     Size = std::min(Size, Word::GetMaxSize());
-    size_t Idx = SimpleFastHash(Data, Size) % kSize;
+    auto Idx = SimpleFastHash(Data, Size) % kSize;
     MemMemWords[Idx].Set(Data, Size);
   }
   const Word &Get(size_t Idx) {
@@ -79,7 +79,7 @@
   void SetPrintNewPCs(bool P) { DoPrintNewPCs = P; }
   void SetPrintNewFuncs(size_t P) { NumPrintNewFuncs = P; }
   void UpdateObservedPCs();
-  template <class Callback> void CollectFeatures(Callback CB) const;
+  template <class Callback> size_t CollectFeatures(Callback CB) const;
 
   void ResetMaps() {
     ValueProfileMap.Reset();
@@ -193,7 +193,7 @@
       Handle8bitCounter(FirstFeature, P - Begin, V);
 
   // Iterate by Step bytes at a time.
-  for (; P < End; P += Step)
+  for (; P + Step <= End; P += Step)
     if (LargeType Bundle = *reinterpret_cast<const LargeType *>(P)) {
       Bundle = HostToLE(Bundle);
       for (size_t I = 0; I < Step; I++, Bundle >>= 8)
@@ -234,16 +234,16 @@
     return Bit;
 }
 
-template <class Callback>  // void Callback(size_t Feature)
-ATTRIBUTE_NO_SANITIZE_ADDRESS
-ATTRIBUTE_NOINLINE
-void TracePC::CollectFeatures(Callback HandleFeature) const {
+template <class Callback> // void Callback(uint32_t Feature)
+ATTRIBUTE_NO_SANITIZE_ADDRESS ATTRIBUTE_NOINLINE size_t
+TracePC::CollectFeatures(Callback HandleFeature) const {
   auto Handle8bitCounter = [&](size_t FirstFeature,
                                size_t Idx, uint8_t Counter) {
     if (UseCounters)
-      HandleFeature(FirstFeature + Idx * 8 + CounterToFeature(Counter));
+      HandleFeature(static_cast<uint32_t>(FirstFeature + Idx * 8 +
+                                          CounterToFeature(Counter)));
     else
-      HandleFeature(FirstFeature + Idx);
+      HandleFeature(static_cast<uint32_t>(FirstFeature + Idx));
   };
 
   size_t FirstFeature = 0;
@@ -263,16 +263,18 @@
 
   if (UseValueProfileMask) {
     ValueProfileMap.ForEach([&](size_t Idx) {
-      HandleFeature(FirstFeature + Idx);
+      HandleFeature(static_cast<uint32_t>(FirstFeature + Idx));
     });
     FirstFeature += ValueProfileMap.SizeInBits();
   }
 
   // Step function, grows similar to 8 * Log_2(A).
-  auto StackDepthStepFunction = [](uint32_t A) -> uint32_t {
-    if (!A) return A;
-    uint32_t Log2 = Log(A);
-    if (Log2 < 3) return A;
+  auto StackDepthStepFunction = [](size_t A) -> size_t {
+    if (!A)
+      return A;
+    auto Log2 = Log(A);
+    if (Log2 < 3)
+      return A;
     Log2 -= 3;
     return (Log2 + 1) * 8 + ((A >> Log2) & 7);
   };
@@ -280,8 +282,13 @@
   assert(StackDepthStepFunction(1024 * 4) == 80);
   assert(StackDepthStepFunction(1024 * 1024) == 144);
 
-  if (auto MaxStackOffset = GetMaxStackOffset())
-    HandleFeature(FirstFeature + StackDepthStepFunction(MaxStackOffset / 8));
+  if (auto MaxStackOffset = GetMaxStackOffset()) {
+    HandleFeature(static_cast<uint32_t>(
+        FirstFeature + StackDepthStepFunction(MaxStackOffset / 8)));
+    FirstFeature += StackDepthStepFunction(std::numeric_limits<size_t>::max());
+  }
+
+  return FirstFeature;
 }
 
 extern TracePC TPC;
diff --git a/libfuzzer/FuzzerUtil.cpp b/libfuzzer/FuzzerUtil.cpp
index 7eecb68..0518549 100644
--- a/libfuzzer/FuzzerUtil.cpp
+++ b/libfuzzer/FuzzerUtil.cpp
@@ -111,7 +111,7 @@
         char Hex[] = "0xAA";
         Hex[2] = Str[Pos + 2];
         Hex[3] = Str[Pos + 3];
-        U->push_back(strtol(Hex, nullptr, 16));
+        U->push_back(static_cast<uint8_t>(strtol(Hex, nullptr, 16)));
         Pos += 3;
         continue;
       }
@@ -226,10 +226,11 @@
   return N;
 }
 
-size_t SimpleFastHash(const uint8_t *Data, size_t Size) {
-  size_t Res = 0;
+uint64_t SimpleFastHash(const void *Data, size_t Size, uint64_t Initial) {
+  uint64_t Res = Initial;
+  const uint8_t *Bytes = static_cast<const uint8_t *>(Data);
   for (size_t i = 0; i < Size; i++)
-    Res = Res * 11 + Data[i];
+    Res = Res * 11 + Bytes[i];
   return Res;
 }
 
diff --git a/libfuzzer/FuzzerUtil.h b/libfuzzer/FuzzerUtil.h
index e90be08..a188a7b 100644
--- a/libfuzzer/FuzzerUtil.h
+++ b/libfuzzer/FuzzerUtil.h
@@ -88,9 +88,11 @@
 
 std::string SearchRegexCmd(const std::string &Regex);
 
-size_t SimpleFastHash(const uint8_t *Data, size_t Size);
+uint64_t SimpleFastHash(const void *Data, size_t Size, uint64_t Initial = 0);
 
-inline uint32_t Log(uint32_t X) { return 32 - Clz(X) - 1; }
+inline size_t Log(size_t X) {
+  return static_cast<size_t>((sizeof(unsigned long long) * 8) - Clzll(X) - 1);
+}
 
 inline size_t PageSize() { return 4096; }
 inline uint8_t *RoundUpByPage(uint8_t *P) {
diff --git a/libfuzzer/FuzzerUtilFuchsia.cpp b/libfuzzer/FuzzerUtilFuchsia.cpp
index af43946..5034b4a 100644
--- a/libfuzzer/FuzzerUtilFuchsia.cpp
+++ b/libfuzzer/FuzzerUtilFuchsia.cpp
@@ -515,7 +515,7 @@
     return rc;
   }
 
-  return Info.return_code;
+  return static_cast<int>(Info.return_code);
 }
 
 bool ExecuteCommand(const Command &BaseCmd, std::string *CmdOutput) {
diff --git a/libfuzzer/FuzzerUtilPosix.cpp b/libfuzzer/FuzzerUtilPosix.cpp
index afb7334..0446d73 100644
--- a/libfuzzer/FuzzerUtilPosix.cpp
+++ b/libfuzzer/FuzzerUtilPosix.cpp
@@ -77,10 +77,13 @@
       return;
   }
 
-  sigact = {};
-  sigact.sa_flags = SA_SIGINFO;
-  sigact.sa_sigaction = callback;
-  if (sigaction(signum, &sigact, 0)) {
+  struct sigaction new_sigact = {};
+  // Address sanitizer needs SA_ONSTACK (causing the signal handler to run on a
+  // dedicated stack) in order to be able to detect stack overflows; keep the
+  // flag if it's set.
+  new_sigact.sa_flags = SA_SIGINFO | (sigact.sa_flags & SA_ONSTACK);
+  new_sigact.sa_sigaction = callback;
+  if (sigaction(signum, &new_sigact, nullptr)) {
     Printf("libFuzzer: sigaction failed with %d\n", errno);
     exit(1);
   }
diff --git a/libfuzzer/tests/FuzzedDataProviderUnittest.cpp b/libfuzzer/tests/FuzzedDataProviderUnittest.cpp
index 99d9d8e..ea6774e 100644
--- a/libfuzzer/tests/FuzzedDataProviderUnittest.cpp
+++ b/libfuzzer/tests/FuzzedDataProviderUnittest.cpp
@@ -283,6 +283,20 @@
   EXPECT_EQ(false, DataProv.ConsumeBool());
 }
 
+TEST(FuzzedDataProvider, PickValueInStdArray) {
+  FuzzedDataProvider DataProv(Data, sizeof(Data));
+  const std::array<int, 5> Array = {1, 2, 3, 4, 5};
+  EXPECT_EQ(5, DataProv.PickValueInArray(Array));
+  EXPECT_EQ(2, DataProv.PickValueInArray(Array));
+  EXPECT_EQ(2, DataProv.PickValueInArray(Array));
+  EXPECT_EQ(3, DataProv.PickValueInArray(Array));
+  EXPECT_EQ(3, DataProv.PickValueInArray(Array));
+  EXPECT_EQ(3, DataProv.PickValueInArray(Array));
+  EXPECT_EQ(1, DataProv.PickValueInArray(Array));
+  EXPECT_EQ(3, DataProv.PickValueInArray(Array));
+  EXPECT_EQ(2, DataProv.PickValueInArray(Array));
+}
+
 TEST(FuzzedDataProvider, PickValueInArray) {
   FuzzedDataProvider DataProv(Data, sizeof(Data));
   const int Array[] = {1, 2, 3, 4, 5};
diff --git a/libfuzzer/tests/FuzzerUnittest.cpp b/libfuzzer/tests/FuzzerUnittest.cpp
index d2b5cbb..974a01f 100644
--- a/libfuzzer/tests/FuzzerUnittest.cpp
+++ b/libfuzzer/tests/FuzzerUnittest.cpp
@@ -614,73 +614,80 @@
   }
 }
 
-TEST(Merge, Bad) {
-  const char *kInvalidInputs[] = {
-    "",
-    "x",
-    "3\nx",
-    "2\n3",
-    "2\n2",
-    "2\n2\nA\n",
-    "2\n2\nA\nB\nC\n",
-    "0\n0\n",
-    "1\n1\nA\nFT 0",
-    "1\n1\nA\nSTARTED 1",
-  };
-  Merger M;
-  for (auto S : kInvalidInputs) {
-    // fprintf(stderr, "TESTING:\n%s\n", S);
-    EXPECT_FALSE(M.Parse(S, false));
-  }
-}
-
-void EQ(const Vector<uint32_t> &A, const Vector<uint32_t> &B) {
+template <typename T> void EQ(const Vector<T> &A, const Vector<T> &B) {
   EXPECT_EQ(A, B);
 }
 
-void EQ(const Vector<std::string> &A, const Vector<std::string> &B) {
-  Set<std::string> a(A.begin(), A.end());
+template <typename T> void EQ(const Set<T> &A, const Vector<T> &B) {
+  EXPECT_EQ(A, Set<T>(B.begin(), B.end()));
+}
+
+void EQ(const Vector<MergeFileInfo> &A, const Vector<std::string> &B) {
+  Set<std::string> a;
+  for (const auto &File : A)
+    a.insert(File.Name);
   Set<std::string> b(B.begin(), B.end());
   EXPECT_EQ(a, b);
 }
 
-static void Merge(const std::string &Input,
-                  const Vector<std::string> Result,
-                  size_t NumNewFeatures) {
-  Merger M;
-  Vector<std::string> NewFiles;
-  Set<uint32_t> NewFeatures, NewCov;
-  EXPECT_TRUE(M.Parse(Input, true));
-  EXPECT_EQ(NumNewFeatures, M.Merge({}, &NewFeatures, {}, &NewCov, &NewFiles));
-  EQ(NewFiles, Result);
-}
+#define TRACED_EQ(A, ...)                                                      \
+  {                                                                            \
+    SCOPED_TRACE(#A);                                                          \
+    EQ(A, __VA_ARGS__);                                                        \
+  }
 
-TEST(Merge, Good) {
+TEST(Merger, Parse) {
   Merger M;
 
+  const char *kInvalidInputs[] = {
+      // Bad file numbers
+      "",
+      "x",
+      "0\n0",
+      "3\nx",
+      "2\n3",
+      "2\n2",
+      // Bad file names
+      "2\n2\nA\n",
+      "2\n2\nA\nB\nC\n",
+      // Unknown markers
+      "2\n1\nA\nSTARTED 0\nBAD 0 0x0",
+      // Bad file IDs
+      "1\n1\nA\nSTARTED 1",
+      "2\n1\nA\nSTARTED 0\nFT 1 0x0",
+  };
+  for (auto S : kInvalidInputs) {
+    SCOPED_TRACE(S);
+    EXPECT_FALSE(M.Parse(S, false));
+  }
+
+  // Parse initial control file
   EXPECT_TRUE(M.Parse("1\n0\nAA\n", false));
-  EXPECT_EQ(M.Files.size(), 1U);
+  ASSERT_EQ(M.Files.size(), 1U);
   EXPECT_EQ(M.NumFilesInFirstCorpus, 0U);
   EXPECT_EQ(M.Files[0].Name, "AA");
   EXPECT_TRUE(M.LastFailure.empty());
   EXPECT_EQ(M.FirstNotProcessedFile, 0U);
 
+  // Parse control file that failed on first attempt
   EXPECT_TRUE(M.Parse("2\n1\nAA\nBB\nSTARTED 0 42\n", false));
-  EXPECT_EQ(M.Files.size(), 2U);
+  ASSERT_EQ(M.Files.size(), 2U);
   EXPECT_EQ(M.NumFilesInFirstCorpus, 1U);
   EXPECT_EQ(M.Files[0].Name, "AA");
   EXPECT_EQ(M.Files[1].Name, "BB");
   EXPECT_EQ(M.LastFailure, "AA");
   EXPECT_EQ(M.FirstNotProcessedFile, 1U);
 
+  // Parse control file that failed on later attempt
   EXPECT_TRUE(M.Parse("3\n1\nAA\nBB\nC\n"
-                        "STARTED 0 1000\n"
-                        "FT 0 1 2 3\n"
-                        "STARTED 1 1001\n"
-                        "FT 1 4 5 6 \n"
-                        "STARTED 2 1002\n"
-                        "", true));
-  EXPECT_EQ(M.Files.size(), 3U);
+                      "STARTED 0 1000\n"
+                      "FT 0 1 2 3\n"
+                      "STARTED 1 1001\n"
+                      "FT 1 4 5 6 \n"
+                      "STARTED 2 1002\n"
+                      "",
+                      true));
+  ASSERT_EQ(M.Files.size(), 3U);
   EXPECT_EQ(M.NumFilesInFirstCorpus, 1U);
   EXPECT_EQ(M.Files[0].Name, "AA");
   EXPECT_EQ(M.Files[0].Size, 1000U);
@@ -690,82 +697,171 @@
   EXPECT_EQ(M.Files[2].Size, 1002U);
   EXPECT_EQ(M.LastFailure, "C");
   EXPECT_EQ(M.FirstNotProcessedFile, 3U);
-  EQ(M.Files[0].Features, {1, 2, 3});
-  EQ(M.Files[1].Features, {4, 5, 6});
+  TRACED_EQ(M.Files[0].Features, {1, 2, 3});
+  TRACED_EQ(M.Files[1].Features, {4, 5, 6});
 
+  // Parse control file without features or PCs
+  EXPECT_TRUE(M.Parse("2\n0\nAA\nBB\n"
+                      "STARTED 0 1000\n"
+                      "FT 0\n"
+                      "COV 0\n"
+                      "STARTED 1 1001\n"
+                      "FT 1\n"
+                      "COV 1\n"
+                      "",
+                      true));
+  ASSERT_EQ(M.Files.size(), 2U);
+  EXPECT_EQ(M.NumFilesInFirstCorpus, 0U);
+  EXPECT_TRUE(M.LastFailure.empty());
+  EXPECT_EQ(M.FirstNotProcessedFile, 2U);
+  EXPECT_TRUE(M.Files[0].Features.empty());
+  EXPECT_TRUE(M.Files[0].Cov.empty());
+  EXPECT_TRUE(M.Files[1].Features.empty());
+  EXPECT_TRUE(M.Files[1].Cov.empty());
 
-  Vector<std::string> NewFiles;
-  Set<uint32_t> NewFeatures, NewCov;
-
+  // Parse features and PCs
   EXPECT_TRUE(M.Parse("3\n2\nAA\nBB\nC\n"
-                        "STARTED 0 1000\nFT 0 1 2 3\n"
-                        "STARTED 1 1001\nFT 1 4 5 6 \n"
-                        "STARTED 2 1002\nFT 2 6 1 3 \n"
-                        "", true));
-  EXPECT_EQ(M.Files.size(), 3U);
+                      "STARTED 0 1000\n"
+                      "FT 0 1 2 3\n"
+                      "COV 0 11 12 13\n"
+                      "STARTED 1 1001\n"
+                      "FT 1 4 5 6\n"
+                      "COV 1 7 8 9\n"
+                      "STARTED 2 1002\n"
+                      "FT 2 6 1 3\n"
+                      "COV 2 16 11 13\n"
+                      "",
+                      true));
+  ASSERT_EQ(M.Files.size(), 3U);
   EXPECT_EQ(M.NumFilesInFirstCorpus, 2U);
   EXPECT_TRUE(M.LastFailure.empty());
   EXPECT_EQ(M.FirstNotProcessedFile, 3U);
-  EQ(M.Files[0].Features, {1, 2, 3});
-  EQ(M.Files[1].Features, {4, 5, 6});
-  EQ(M.Files[2].Features, {1, 3, 6});
-  EXPECT_EQ(0U, M.Merge({}, &NewFeatures, {}, &NewCov, &NewFiles));
-  EQ(NewFiles, {});
+  TRACED_EQ(M.Files[0].Features, {1, 2, 3});
+  TRACED_EQ(M.Files[0].Cov, {11, 12, 13});
+  TRACED_EQ(M.Files[1].Features, {4, 5, 6});
+  TRACED_EQ(M.Files[1].Cov, {7, 8, 9});
+  TRACED_EQ(M.Files[2].Features, {1, 3, 6});
+  TRACED_EQ(M.Files[2].Cov, {16});
+}
 
+TEST(Merger, Merge) {
+  Merger M;
+  Set<uint32_t> Features, NewFeatures;
+  Set<uint32_t> Cov, NewCov;
+  Vector<std::string> NewFiles;
+
+  // Adds new files and features
+  EXPECT_TRUE(M.Parse("3\n0\nA\nB\nC\n"
+                      "STARTED 0 1000\n"
+                      "FT 0 1 2 3\n"
+                      "STARTED 1 1001\n"
+                      "FT 1 4 5 6 \n"
+                      "STARTED 2 1002\n"
+                      "FT 2 6 1 3\n"
+                      "",
+                      true));
+  EXPECT_EQ(M.Merge(Features, &NewFeatures, Cov, &NewCov, &NewFiles), 6U);
+  TRACED_EQ(M.Files, {"A", "B", "C"});
+  TRACED_EQ(NewFiles, {"A", "B"});
+  TRACED_EQ(NewFeatures, {1, 2, 3, 4, 5, 6});
+
+  // Doesn't return features or files in the initial corpus.
   EXPECT_TRUE(M.Parse("3\n1\nA\nB\nC\n"
-                        "STARTED 0 1000\nFT 0 1 2 3\n"
-                        "STARTED 1 1001\nFT 1 4 5 6 \n"
-                        "STARTED 2 1002\nFT 2 6 1 3\n"
-                        "", true));
-  EQ(M.Files[0].Features, {1, 2, 3});
-  EQ(M.Files[1].Features, {4, 5, 6});
-  EQ(M.Files[2].Features, {1, 3, 6});
-  EXPECT_EQ(3U, M.Merge({}, &NewFeatures, {}, &NewCov, &NewFiles));
-  EQ(NewFiles, {"B"});
+                      "STARTED 0 1000\n"
+                      "FT 0 1 2 3\n"
+                      "STARTED 1 1001\n"
+                      "FT 1 4 5 6 \n"
+                      "STARTED 2 1002\n"
+                      "FT 2 6 1 3\n"
+                      "",
+                      true));
+  EXPECT_EQ(M.Merge(Features, &NewFeatures, Cov, &NewCov, &NewFiles), 3U);
+  TRACED_EQ(M.Files, {"A", "B", "C"});
+  TRACED_EQ(NewFiles, {"B"});
+  TRACED_EQ(NewFeatures, {4, 5, 6});
 
-  // Same as the above, but with InitialFeatures.
-  EXPECT_TRUE(M.Parse("2\n0\nB\nC\n"
-                        "STARTED 0 1001\nFT 0 4 5 6 \n"
-                        "STARTED 1 1002\nFT 1 6 1 3\n"
-                        "", true));
-  EQ(M.Files[0].Features, {4, 5, 6});
-  EQ(M.Files[1].Features, {1, 3, 6});
-  Set<uint32_t> InitialFeatures;
-  InitialFeatures.insert(1);
-  InitialFeatures.insert(2);
-  InitialFeatures.insert(3);
-  EXPECT_EQ(3U, M.Merge(InitialFeatures, &NewFeatures, {}, &NewCov, &NewFiles));
-  EQ(NewFiles, {"B"});
+  // No new features, so no new files
+  EXPECT_TRUE(M.Parse("3\n2\nA\nB\nC\n"
+                      "STARTED 0 1000\n"
+                      "FT 0 1 2 3\n"
+                      "STARTED 1 1001\n"
+                      "FT 1 4 5 6 \n"
+                      "STARTED 2 1002\n"
+                      "FT 2 6 1 3\n"
+                      "",
+                      true));
+  EXPECT_EQ(M.Merge(Features, &NewFeatures, Cov, &NewCov, &NewFiles), 0U);
+  TRACED_EQ(M.Files, {"A", "B", "C"});
+  TRACED_EQ(NewFiles, {});
+  TRACED_EQ(NewFeatures, {});
+
+  // Can pass initial features and coverage.
+  Features = {1, 2, 3};
+  Cov = {};
+  EXPECT_TRUE(M.Parse("2\n0\nA\nB\n"
+                      "STARTED 0 1000\n"
+                      "FT 0 1 2 3\n"
+                      "STARTED 1 1001\n"
+                      "FT 1 4 5 6\n"
+                      "",
+                      true));
+  EXPECT_EQ(M.Merge(Features, &NewFeatures, Cov, &NewCov, &NewFiles), 3U);
+  TRACED_EQ(M.Files, {"A", "B"});
+  TRACED_EQ(NewFiles, {"B"});
+  TRACED_EQ(NewFeatures, {4, 5, 6});
+  Features.clear();
+  Cov.clear();
+
+  // Parse smaller files first
+  EXPECT_TRUE(M.Parse("3\n0\nA\nB\nC\n"
+                      "STARTED 0 2000\n"
+                      "FT 0 1 2 3\n"
+                      "STARTED 1 1001\n"
+                      "FT 1 4 5 6 \n"
+                      "STARTED 2 1002\n"
+                      "FT 2 6 1 3 \n"
+                      "",
+                      true));
+  EXPECT_EQ(M.Merge(Features, &NewFeatures, Cov, &NewCov, &NewFiles), 6U);
+  TRACED_EQ(M.Files, {"B", "C", "A"});
+  TRACED_EQ(NewFiles, {"B", "C", "A"});
+  TRACED_EQ(NewFeatures, {1, 2, 3, 4, 5, 6});
+
+  EXPECT_TRUE(M.Parse("4\n0\nA\nB\nC\nD\n"
+                      "STARTED 0 2000\n"
+                      "FT 0 1 2 3\n"
+                      "STARTED 1 1101\n"
+                      "FT 1 4 5 6 \n"
+                      "STARTED 2 1102\n"
+                      "FT 2 6 1 3 100 \n"
+                      "STARTED 3 1000\n"
+                      "FT 3 1  \n"
+                      "",
+                      true));
+  EXPECT_EQ(M.Merge(Features, &NewFeatures, Cov, &NewCov, &NewFiles), 7U);
+  TRACED_EQ(M.Files, {"A", "B", "C", "D"});
+  TRACED_EQ(NewFiles, {"D", "B", "C", "A"});
+  TRACED_EQ(NewFeatures, {1, 2, 3, 4, 5, 6, 100});
+
+  // For same sized file, parse more features first
+  EXPECT_TRUE(M.Parse("4\n1\nA\nB\nC\nD\n"
+                      "STARTED 0 2000\n"
+                      "FT 0 4 5 6 7 8\n"
+                      "STARTED 1 1100\n"
+                      "FT 1 1 2 3 \n"
+                      "STARTED 2 1100\n"
+                      "FT 2 2 3 \n"
+                      "STARTED 3 1000\n"
+                      "FT 3 1  \n"
+                      "",
+                      true));
+  EXPECT_EQ(M.Merge(Features, &NewFeatures, Cov, &NewCov, &NewFiles), 3U);
+  TRACED_EQ(M.Files, {"A", "B", "C", "D"});
+  TRACED_EQ(NewFiles, {"D", "B"});
+  TRACED_EQ(NewFeatures, {1, 2, 3});
 }
 
-TEST(Merge, Merge) {
-
-  Merge("3\n1\nA\nB\nC\n"
-        "STARTED 0 1000\nFT 0 1 2 3\n"
-        "STARTED 1 1001\nFT 1 4 5 6 \n"
-        "STARTED 2 1002\nFT 2 6 1 3 \n",
-        {"B"}, 3);
-
-  Merge("3\n0\nA\nB\nC\n"
-        "STARTED 0 2000\nFT 0 1 2 3\n"
-        "STARTED 1 1001\nFT 1 4 5 6 \n"
-        "STARTED 2 1002\nFT 2 6 1 3 \n",
-        {"A", "B", "C"}, 6);
-
-  Merge("4\n0\nA\nB\nC\nD\n"
-        "STARTED 0 2000\nFT 0 1 2 3\n"
-        "STARTED 1 1101\nFT 1 4 5 6 \n"
-        "STARTED 2 1102\nFT 2 6 1 3 100 \n"
-        "STARTED 3 1000\nFT 3 1  \n",
-        {"A", "B", "C", "D"}, 7);
-
-  Merge("4\n1\nA\nB\nC\nD\n"
-        "STARTED 0 2000\nFT 0 4 5 6 7 8\n"
-        "STARTED 1 1100\nFT 1 1 2 3 \n"
-        "STARTED 2 1100\nFT 2 2 3 \n"
-        "STARTED 3 1000\nFT 3 1  \n",
-        {"B", "D"}, 3);
-}
+#undef TRACED_EQ
 
 TEST(DFT, BlockCoverage) {
   BlockCoverage Cov;
diff --git a/patches/Android.bp.patch b/patches/Android.bp.patch
deleted file mode 100644
index c0d6fd7..0000000
--- a/patches/Android.bp.patch
+++ /dev/null
@@ -1,25 +0,0 @@
-diff --git a/Android.bp b/Android.bp
-index 86434dd..e565b86 100644
---- a/Android.bp
-+++ b/Android.bp
-@@ -1,17 +1,15 @@
- // This file is generated by cargo2android.py --run --dependencies --device --features arbitrary-derive --patch patches/Android.bp.patch.
- 
--rust_library {
-+rust_library_rlib {
-     name: "liblibfuzzer_sys",
-     host_supported: true,
-     crate_name: "libfuzzer_sys",
-     srcs: ["src/lib.rs"],
-     edition: "2018",
-     features: ["arbitrary-derive"],
--    rustlibs: [
-+    rlibs: [
-         "libarbitrary",
-     ],
--    static_libs: ["libfuzzer"],
--    shared_libs: ["libstdc++"],
- }
- 
- // dependent_library ["feature_list"]
-
diff --git a/src/lib.rs b/src/lib.rs
index 59dfc47..6badc39 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -12,11 +12,14 @@
 #![deny(missing_docs, missing_debug_implementations)]
 
 pub use arbitrary;
+use once_cell::sync::OnceCell;
 
 extern "C" {
     // We do not actually cross the FFI bound here.
     #[allow(improper_ctypes)]
     fn rust_fuzzer_test_input(input: &[u8]);
+
+    fn LLVMFuzzerMutate(data: *mut u8, size: usize, max_size: usize) -> usize;
 }
 
 #[doc(hidden)]
@@ -35,6 +38,9 @@
 }
 
 #[doc(hidden)]
+pub static RUST_LIBFUZZER_DEBUG_PATH: OnceCell<String> = OnceCell::new();
+
+#[doc(hidden)]
 #[export_name = "LLVMFuzzerInitialize"]
 pub fn initialize(_argc: *const isize, _argv: *const *const *const u8) -> isize {
     // Registers a panic hook that aborts the process before unwinding.
@@ -50,6 +56,14 @@
         default_hook(panic_info);
         ::std::process::abort();
     }));
+
+    // Initialize the `RUST_LIBFUZZER_DEBUG_PATH` cell with the path so it can be
+    // reused with little overhead.
+    if let Ok(path) = std::env::var("RUST_LIBFUZZER_DEBUG_PATH") {
+        RUST_LIBFUZZER_DEBUG_PATH
+            .set(path)
+            .expect("Since this is initialize it is only called once so can never fail");
+    }
     0
 }
 
@@ -84,8 +98,9 @@
 ///
 /// ```no_run
 /// #![no_main]
+/// # mod foo {
 ///
-/// use libfuzzer_sys::{arbitrary::{Arbitrary, Unstructured}, fuzz_target};
+/// use libfuzzer_sys::{arbitrary::{Arbitrary, Error, Unstructured}, fuzz_target};
 ///
 /// #[derive(Debug)]
 /// pub struct Rgb {
@@ -94,11 +109,8 @@
 ///     b: u8,
 /// }
 ///
-/// impl Arbitrary for Rgb {
-///     fn arbitrary<U>(raw: &mut U) -> Result<Self, U::Error>
-///     where
-///         U: Unstructured + ?Sized
-///     {
+/// impl<'a> Arbitrary<'a> for Rgb {
+///     fn arbitrary(raw: &mut Unstructured<'a>) -> Result<Self, Error> {
 ///         let mut buf = [0; 3];
 ///         raw.fill_buffer(&mut buf)?;
 ///         let r = buf[0];
@@ -112,16 +124,27 @@
 /// fuzz_target!(|color: Rgb| {
 ///     my_crate::convert_color(color);
 /// });
-/// # mod my_crate { fn convert_color(_: super::Rgb) {} }
+/// # mod my_crate {
+/// #     use super::Rgb;
+/// #     pub fn convert_color(_: Rgb) {}
+/// # }
+/// # }
+/// ```
+///
+/// You can also enable the `arbitrary` crate's custom derive via this crate's
+/// `"arbitrary-derive"` cargo feature.
 #[macro_export]
 macro_rules! fuzz_target {
     (|$bytes:ident| $body:block) => {
+        /// Auto-generated function
         #[no_mangle]
         pub extern "C" fn rust_fuzzer_test_input($bytes: &[u8]) {
             // When `RUST_LIBFUZZER_DEBUG_PATH` is set, write the debug
             // formatting of the input to that file. This is only intended for
             // `cargo fuzz`'s use!
-            if let Ok(path) = std::env::var("RUST_LIBFUZZER_DEBUG_PATH") {
+
+            // `RUST_LIBFUZZER_DEBUG_PATH` is set in initialization.
+            if let Some(path) = $crate::RUST_LIBFUZZER_DEBUG_PATH.get() {
                 use std::io::Write;
                 let mut file = std::fs::File::create(path)
                     .expect("failed to create `RUST_LIBFUZZER_DEBUG_PATH` file");
@@ -139,9 +162,10 @@
     };
 
     (|$data:ident: $dty: ty| $body:block) => {
+        /// Auto-generated function
         #[no_mangle]
         pub extern "C" fn rust_fuzzer_test_input(bytes: &[u8]) {
-            use libfuzzer_sys::arbitrary::{Arbitrary, Unstructured};
+            use $crate::arbitrary::{Arbitrary, Unstructured};
 
             // Early exit if we don't have enough bytes for the `Arbitrary`
             // implementation. This helps the fuzzer avoid exploring all the
@@ -159,7 +183,9 @@
             // When `RUST_LIBFUZZER_DEBUG_PATH` is set, write the debug
             // formatting of the input to that file. This is only intended for
             // `cargo fuzz`'s use!
-            if let Ok(path) = std::env::var("RUST_LIBFUZZER_DEBUG_PATH") {
+
+            // `RUST_LIBFUZZER_DEBUG_PATH` is set in initialization.
+            if let Some(path) = $crate::RUST_LIBFUZZER_DEBUG_PATH.get() {
                 use std::io::Write;
                 let mut file = std::fs::File::create(path)
                     .expect("failed to create `RUST_LIBFUZZER_DEBUG_PATH` file");
@@ -180,3 +206,221 @@
         }
     };
 }
+
+/// Define a custom mutator.
+///
+/// This is optional, and libFuzzer will use its own, default mutation strategy
+/// if this is not provided.
+///
+/// You might consider using a custom mutator when your fuzz target is very
+/// particular about the shape of its input:
+///
+/// * You want to fuzz "deeper" than just the parser.
+/// * The input contains checksums that have to match the hash of some subset of
+///   the data or else the whole thing is invalid, and therefore mutating any of
+///   that subset means you need to recompute the checksums.
+/// * Small random changes to the input buffer make it invalid.
+///
+/// That is, a custom mutator is useful in similar situations where [a `T:
+/// Arbitrary` input type](macro.fuzz_target.html#arbitrary-input-types) is
+/// useful. Note that the two approaches are not mutually exclusive; you can use
+/// whichever is easier for your problem domain or both!
+///
+/// ## Implementation Contract
+///
+/// The original, unmodified input is given in `data[..size]`.
+///
+/// You must modify the data in place and return the new size.
+///
+/// The new size should not be greater than `max_size`. If this is not the case,
+/// then the `data` will be truncated to fit within `max_size`. Note that
+/// `max_size < size` is possible when shrinking test cases.
+///
+/// You must produce the same mutation given the same `seed`. Generally, when
+/// choosing what kind of mutation to make or where to mutate, you should start
+/// by creating a random number generator (RNG) that is seeded with the given
+/// `seed` and then consult the RNG whenever making a decision:
+///
+/// ```no_run
+/// #![no_main]
+///
+/// use rand::{rngs::StdRng, Rng, SeedableRng};
+///
+/// libfuzzer_sys::fuzz_mutator!(|data: &mut [u8], size: usize, max_size: usize, seed: u32| {
+///     let mut rng = StdRng::seed_from_u64(seed as u64);
+///
+/// #   let first_mutation = |_, _, _, _| todo!();
+/// #   let second_mutation = |_, _, _, _| todo!();
+/// #   let third_mutation = |_, _, _, _| todo!();
+/// #   let fourth_mutation = |_, _, _, _| todo!();
+///     // Choose which of our four supported kinds of mutations we want to make.
+///     match rng.gen_range(0..4) {
+///         0 => first_mutation(rng, data, size, max_size),
+///         1 => second_mutation(rng, data, size, max_size),
+///         2 => third_mutation(rng, data, size, max_size),
+///         3 => fourth_mutation(rng, data, size, max_size),
+///         _ => unreachable!()
+///     }
+/// });
+/// ```
+///
+/// ## Example: Compression
+///
+/// Consider a simple fuzz target that takes compressed data as input,
+/// decompresses it, and then asserts that the decompressed data doesn't begin
+/// with "boom". It is difficult for `libFuzzer` (or any other fuzzer) to crash
+/// this fuzz target because nearly all mutations it makes will invalidate the
+/// compression format. Therefore, we use a custom mutator that decompresses the
+/// raw input, mutates the decompressed data, and then recompresses it. This
+/// allows `libFuzzer` to quickly discover crashing inputs.
+///
+/// ```no_run
+/// #![no_main]
+///
+/// use flate2::{read::GzDecoder, write::GzEncoder, Compression};
+/// use libfuzzer_sys::{fuzz_mutator, fuzz_target};
+/// use std::io::{Read, Write};
+///
+/// fuzz_target!(|data: &[u8]| {
+///     // Decompress the input data and crash if it starts with "boom".
+///     if let Some(data) = decompress(data) {
+///         if data.starts_with(b"boom") {
+///             panic!();
+///         }
+///     }
+/// });
+///
+/// fuzz_mutator!(
+///     |data: &mut [u8], size: usize, max_size: usize, _seed: u32| {
+///         // Decompress the input data. If that fails, use a dummy value.
+///         let mut decompressed = decompress(&data[..size]).unwrap_or_else(|| b"hi".to_vec());
+///
+///         // Mutate the decompressed data with `libFuzzer`'s default mutator. Make
+///         // the `decompressed` vec's extra capacity available for insertion
+///         // mutations via `resize`.
+///         let len = decompressed.len();
+///         let cap = decompressed.capacity();
+///         decompressed.resize(cap, 0);
+///         let new_decompressed_size = libfuzzer_sys::fuzzer_mutate(&mut decompressed, len, cap);
+///
+///         // Recompress the mutated data.
+///         let compressed = compress(&decompressed[..new_decompressed_size]);
+///
+///         // Copy the recompressed mutated data into `data` and return the new size.
+///         let new_size = std::cmp::min(max_size, compressed.len());
+///         data[..new_size].copy_from_slice(&compressed[..new_size]);
+///         new_size
+///     }
+/// );
+///
+/// fn decompress(compressed_data: &[u8]) -> Option<Vec<u8>> {
+///     let mut decoder = GzDecoder::new(compressed_data);
+///     let mut decompressed = Vec::new();
+///     if decoder.read_to_end(&mut decompressed).is_ok() {
+///         Some(decompressed)
+///     } else {
+///         None
+///     }
+/// }
+///
+/// fn compress(data: &[u8]) -> Vec<u8> {
+///     let mut encoder = GzEncoder::new(Vec::new(), Compression::default());
+///     encoder
+///         .write_all(data)
+///         .expect("writing into a vec is infallible");
+///     encoder.finish().expect("writing into a vec is infallible")
+/// }
+/// ```
+///
+/// This example is inspired by [a similar example from the official `libFuzzer`
+/// docs](https://github.com/google/fuzzing/blob/master/docs/structure-aware-fuzzing.md#example-compression).
+///
+/// ## More Example Ideas
+///
+/// * A PNG custom mutator that decodes a PNG, mutates the image, and then
+/// re-encodes the mutated image as a new PNG.
+///
+/// * A [`serde`](https://serde.rs/) custom mutator that deserializes your
+///   structure, mutates it, and then reserializes it.
+///
+/// * A Wasm binary custom mutator that inserts, replaces, and removes a
+///   bytecode instruction in a function's body.
+///
+/// * An HTTP request custom mutator that inserts, replaces, and removes a
+///   header from an HTTP request.
+#[macro_export]
+macro_rules! fuzz_mutator {
+    (
+        |
+        $data:ident : &mut [u8] ,
+        $size:ident : usize ,
+        $max_size:ident : usize ,
+        $seed:ident : u32 $(,)*
+        |
+        $body:block
+    ) => {
+        /// Auto-generated function.
+        #[export_name = "LLVMFuzzerCustomMutator"]
+        pub fn rust_fuzzer_custom_mutator(
+            $data: *mut u8,
+            $size: usize,
+            $max_size: usize,
+            $seed: std::os::raw::c_uint,
+        ) -> usize {
+            // Depending on if we are growing or shrinking the test case, `size`
+            // might be larger or smaller than `max_size`. The `data`'s capacity
+            // is the maximum of the two.
+            let len = std::cmp::max($max_size, $size);
+            let $data: &mut [u8] = unsafe { std::slice::from_raw_parts_mut($data, len) };
+
+            // `unsigned int` is generally a `u32`, but not on all targets. Do
+            // an infallible (and potentially lossy, but that's okay because it
+            // preserves determinism) conversion.
+            let $seed = $seed as u32;
+
+            // Truncate the new size if it is larger than the max.
+            let new_size = { $body };
+            std::cmp::min(new_size, $max_size)
+        }
+    };
+}
+
+/// The default `libFuzzer` mutator.
+///
+/// You generally don't have to use this at all unless you're defining a
+/// custom mutator with [the `fuzz_mutator!` macro][crate::fuzz_mutator].
+///
+/// Mutates `data[..size]` in place such that the mutated data is no larger than
+/// `max_size` and returns the new size of the mutated data.
+///
+/// To only allow shrinking mutations, make `max_size < size`.
+///
+/// To additionally allow mutations that grow the size of the data, make
+/// `max_size > size`.
+///
+/// Both `size` and `max_size` must be less than or equal to `data.len()`.
+///
+/// # Example
+///
+/// ```no_run
+/// // Create some data in a buffer.
+/// let mut data = vec![0; 128];
+/// data[..b"hello".len()].copy_from_slice(b"hello");
+///
+/// // Ask `libFuzzer` to mutate the data. By setting `max_size` to our buffer's
+/// // full length, we are allowing `libFuzzer` to perform mutations that grow
+/// // the size of the data, such as insertions.
+/// let size = b"hello".len();
+/// let max_size = data.len();
+/// let new_size = libfuzzer_sys::fuzzer_mutate(&mut data, size, max_size);
+///
+/// // Get the mutated data out of the buffer.
+/// let mutated_data = &data[..new_size];
+/// ```
+pub fn fuzzer_mutate(data: &mut [u8], size: usize, max_size: usize) -> usize {
+    assert!(size <= data.len());
+    assert!(max_size <= data.len());
+    let new_size = unsafe { LLVMFuzzerMutate(data.as_mut_ptr(), size, max_size) };
+    assert!(new_size <= data.len());
+    new_size
+}