diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index b78fd510..0abab908 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -377,8 +377,8 @@ jobs:
           - { zig_target: x86_64-macos.13-none }
           - { zig_target: x86_64-windows-gnu }
     env:
-      # 2023-04-17
-      ZIG_DIST_VERSION: 0.11.0-dev.2624+bc804eb84
+      # 2023-04-24
+      ZIG_DIST_VERSION: 0.11.0-dev.2777+b95cdf0ae
       # for zig-cc wrapper scripts (see below):
       ZIG_CPPFLAGS: -DUPX_DOCTEST_CONFIG_MULTITHREADING
       ZIG_FLAGS: ${{ matrix.zig_flags }}
diff --git a/src/check/dt_check.cpp b/src/check/dt_check.cpp
index cf268f3c..a999cbee 100644
--- a/src/check/dt_check.cpp
+++ b/src/check/dt_check.cpp
@@ -165,6 +165,12 @@ template <class T>
 struct CheckIntegral {
     template <class U>
     static void checkU(void) {
+        U a = {};
+        const U b = {};
+        constexpr U c = {};
+        UNUSED(a);
+        UNUSED(b);
+        UNUSED(c);
 #if __cplusplus < 202002L
         COMPILE_TIME_ASSERT(std::is_pod<U>::value) // deprecated in C++20
 #endif
@@ -217,15 +223,18 @@ struct TestBELE {
     static noinline bool test(void) {
         CheckIntegral<T>::check();
         CheckAlignment<T>::check();
+        // array init
+        T array[2] = {};
+        assert(array[0] == 0 && array[1] == 0);
         // arithmetic checks (modern compilers will optimize this away)
-        T allbits;
-        allbits = 0;
+        T allbits = {};
+        assert(allbits == 0);
         allbits += 1;
         allbits -= 2;
         T v1;
         v1 = 1;
-        v1 *= 2;
-        v1 /= 1;
+        v1 *= 4;
+        v1 /= 2;
         v1 -= 1;
         T v2;
         v2 = 1;
@@ -251,6 +260,8 @@ struct TestBELE {
         v2 &= v1;
         v2 /= v1;
         v2 *= v1;
+        v1 += v2;
+        v1 -= v2;
         assert(v1 == 1);
         assert(v2 == 0);
         if ((v1 ^ v2) != 1)
diff --git a/src/conf.h b/src/conf.h
index dd4f0c0a..711a3659 100644
--- a/src/conf.h
+++ b/src/conf.h
@@ -65,14 +65,8 @@ ACC_COMPILE_TIME_ASSERT_HEADER((char)(-1) == 255) // -funsigned-char
    //   51:32: error: zero as null pointer constant
 #  pragma GCC diagnostic error "-Wzero-as-null-pointer-constant"
 #endif
-#if (ACC_CC_MSC)
-#  pragma warning(error: 4127)
-#  pragma warning(error: 4146)
-#  pragma warning(error: 4319)
-#  pragma warning(error: 4805)
-#endif
-#endif // UPX_CONFIG_DISABLE_WSTRICT
 #endif // UPX_CONFIG_DISABLE_WERROR
+#endif // UPX_CONFIG_DISABLE_WSTRICT
 
 // multithreading (UPX currently does not use multithreading)
 #if (WITH_THREADS)
@@ -346,7 +340,7 @@ inline void NO_fprintf(FILE *, const char *, ...) {}
 #  define upx_return_address()  nullptr
 #endif
 
-// TODO cleanup: we now require C++14, so remove all __packed_struct usage
+// TODO cleanup: we now require C++17, so remove all __packed_struct usage
 #define __packed_struct(s)      struct alignas(1) s {
 #define __packed_struct_end()   };
 
diff --git a/src/except.h b/src/except.h
index 5fc8cb2a..7713ba05 100644
--- a/src/except.h
+++ b/src/except.h
@@ -211,6 +211,7 @@ NORET void throwOutOfMemoryException(const char *msg = nullptr);
 NORET void throwIOException(const char *msg = nullptr, int e = 0);
 NORET void throwEOFException(const char *msg = nullptr, int e = 0);
 
+// some C++ template wizardry is needed to overload throwCantPack() for varargs
 template <class T>
 void throwCantPack(const T *, ...) = delete;
 template <>
diff --git a/src/main.cpp b/src/main.cpp
index b0c99961..0a749df1 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -28,9 +28,8 @@
 // main entry, mostly boring stuff; see work.cpp for actual action
 
 #include "conf.h"
-#include "file.h"
-#include "packer.h"
-#include "p_elf.h"
+#include "packer.h"            // Packer::isValidCompressionMethod()
+#include "p_elf.h"             // ELFOSABI_xxx
 #include "compress/compress.h" // upx_ucl_init()
 
 /*************************************************************************
diff --git a/src/packer.h b/src/packer.h
index 0102df0c..9c36cda3 100644
--- a/src/packer.h
+++ b/src/packer.h
@@ -278,8 +278,16 @@ protected:
                                     unsigned image_size, int bits, bool bswap);
 
     // Target Endianness abstraction
-#if 1
-    // try to detect TE16 vs TE32 vs TE64 size mismatches; note that "byte" is explicitly allowed
+#if 0
+    // permissive version using "void *"
+    inline unsigned get_te16(const void *p) const noexcept { return bele->get16(p); }
+    inline unsigned get_te32(const void *p) const noexcept { return bele->get32(p); }
+    inline upx_uint64_t get_te64(const void *p) const noexcept { return bele->get64(p); }
+    inline void set_te16(void *p, unsigned v) noexcept { bele->set16(p, v); }
+    inline void set_te32(void *p, unsigned v) noexcept { bele->set32(p, v); }
+    inline void set_te64(void *p, upx_uint64_t v) noexcept { bele->set64(p, v); }
+#else
+    // try to detect TE16 vs TE32 vs TE64 size mismatches; note that byte is explicitly allowed
     template <class T>
     static inline constexpr bool is_te16_type = is_same_any_v<T, byte, upx_uint16_t, BE16, LE16>;
     template <class T>
@@ -312,14 +320,6 @@ protected:
     inline void set_te64(T *p, upx_uint64_t v) noexcept {
         bele->set64(p, v);
     }
-#else
-    // permissive version using "void *"
-    inline unsigned get_te16(const void *p) const noexcept { return bele->get16(p); }
-    inline unsigned get_te32(const void *p) const noexcept { return bele->get32(p); }
-    inline upx_uint64_t get_te64(const void *p) const noexcept { return bele->get64(p); }
-    inline void set_te16(void *p, unsigned v) noexcept { bele->set16(p, v); }
-    inline void set_te32(void *p, unsigned v) noexcept { bele->set32(p, v); }
-    inline void set_te64(void *p, upx_uint64_t v) noexcept { bele->set64(p, v); }
 #endif
 
 protected:
diff --git a/src/util/util.cpp b/src/util/util.cpp
index cd21c0ce..7c19324f 100644
--- a/src/util/util.cpp
+++ b/src/util/util.cpp
@@ -247,20 +247,6 @@ TEST_CASE("ptr_check_no_overlap 3") {
 }
 #endif // DEBUG
 
-/*************************************************************************
-// bele.h
-**************************************************************************/
-
-namespace N_BELE_CTP {
-const BEPolicy be_policy;
-const LEPolicy le_policy;
-} // namespace N_BELE_CTP
-
-namespace N_BELE_RTP {
-const BEPolicy be_policy;
-const LEPolicy le_policy;
-} // namespace N_BELE_RTP
-
 /*************************************************************************
 // stdlib
 **************************************************************************/
@@ -312,25 +298,47 @@ TEST_CASE("upx_stable_sort") {
         CHECK((a[0] == 0 && a[1] == 1));
     }
     {
-        LE64 a[3];
-        a[0] = 2;
-        a[1] = 1;
-        a[2] = 0;
-        upx_stable_sort(a, 3, sizeof(*a), le64_compare);
-        CHECK((a[0] == 0 && a[1] == 1 && a[2] == 2));
+        BE64 a[3];
+        a[0] = 257;
+        a[1] = 256;
+        a[2] = 255;
+        upx_stable_sort(a, 3, sizeof(*a), be64_compare);
+        CHECK((a[0] == 255 && a[1] == 256 && a[2] == 257));
     }
-#if __cplusplus >= 202002L // use C++20 std::next_permutation() to test all permutations
-    {
-        upx_uint16_t perm[5] = {0, 1, 2, 3, 4}; // 120 permutations
-        do {
-            upx_uint16_t a[5] = {};
-            memcpy(a, perm, sizeof(perm));
-            upx_stable_sort(a, 5, sizeof(*a), ne16_compare);
-            CHECK((a[0] == 0 && a[1] == 1 && a[2] == 2 && a[3] == 3 && a[4] == 4));
-        } while (std::next_permutation(perm, perm + 5));
-    }
-#endif
 }
+
+#if __cplusplus >= 202002L // use C++20 std::next_permutation() to test all permutations
+namespace {
+struct TestSortAllPermutations {
+    static upx_uint64_t test(size_t n) {
+        constexpr size_t N = 16;
+        assert(n > 0 && n <= N);
+        LE16 perm[N];
+        for (size_t i = 0; i < n; i++)
+            perm[i] = 255 + i;
+        upx_uint64_t num_perms = 0;
+        do {
+            LE16 a[N];
+            memcpy(a, perm, sizeof(*a) * n);
+            upx_stable_sort(a, n, sizeof(*a), le16_compare);
+            for (size_t i = 0; i < n; i++)
+                CHECK((a[i] == 255 + i));
+            num_perms += 1;
+        } while (std::next_permutation(perm, perm + n));
+        return num_perms;
+    }
+};
+} // namespace
+TEST_CASE("upx_stable_sort") {
+    CHECK(TestSortAllPermutations::test(1) == 1);
+    CHECK(TestSortAllPermutations::test(2) == 2);
+    CHECK(TestSortAllPermutations::test(3) == 6);
+    CHECK(TestSortAllPermutations::test(4) == 24);
+    CHECK(TestSortAllPermutations::test(5) == 120);
+    // CHECK(TestSortAllPermutations::test(6) == 720);
+    // CHECK(TestSortAllPermutations::test(7) == 5040);
+}
+#endif // C++20
 #endif // DEBUG
 
 /*************************************************************************
@@ -537,6 +545,20 @@ TEST_CASE("mem_replace") {
     CHECK(strcmp(b, "cdcdcdcdefgefghh") == 0);
 }
 
+/*************************************************************************
+// bele.h globals
+**************************************************************************/
+
+namespace N_BELE_CTP {
+const BEPolicy be_policy;
+const LEPolicy le_policy;
+} // namespace N_BELE_CTP
+
+namespace N_BELE_RTP {
+const BEPolicy be_policy;
+const LEPolicy le_policy;
+} // namespace N_BELE_RTP
+
 /*************************************************************************
 // fn - FileName util
 **************************************************************************/
@@ -736,7 +758,7 @@ bool makebakname(char *ofilename, size_t size, const char *ifilename, bool force
 **************************************************************************/
 
 unsigned get_ratio(upx_uint64_t u_len, upx_uint64_t c_len) {
-    const unsigned n = 1000 * 1000;
+    constexpr unsigned n = 1000 * 1000;
     if (u_len == 0)
         return c_len == 0 ? 0 : n;
     upx_uint64_t x = c_len * n;