CI updates

.github/workflows/ci.yml

@@ -609,6 +609,8 @@ jobs:
           - { zig_target: powerpc64-linux-musl, qemu: qemu-ppc64, zig_pic: -fPIE }
           - { zig_target: powerpc64le-linux-musl, qemu: qemu-ppc64le }
           - { zig_target: powerpc64le-linux-musl, qemu: qemu-ppc64le, zig_pic: -fPIE }
+          - { zig_target: riscv64-linux-musl, qemu: qemu-riscv64, tt: UPX-UNSUPPORTED }
+          - { zig_target: riscv64-linux-musl, qemu: qemu-riscv64, tt: UPX-UNSUPPORTED, zig_pic: -fPIE }
           - { zig_target: x86_64-linux-gnu.2.3.4, qemu: qemu-x86_64 } # can use QEMU because of gcompat
           - { zig_target: x86_64-linux-gnu.2.17, qemu: qemu-x86_64 }  # can use QEMU because of gcompat
           - { zig_target: x86_64-linux-musl, qemu: qemu-x86_64 }
@@ -652,6 +654,7 @@ jobs:
             # TODO FIXME: problem with self-packed upx and musl+gcompat: "Not a valid dynamic program"
             echo "UPX_CONFIG_DISABLE_RUN_PACKED_TEST=ON" >> $GITHUB_ENV
           fi
+          test -z "${{ matrix.tt }}" || echo "UPX_CONFIG_DISABLE_SELF_PACK_TEST=ON" >> $GITHUB_ENV
           # install zig; note that ~/.local/bin is included in the default $PATH on Ubuntu
           mkdir -p -v ~/.local/bin
           cd ~/.local/bin

CMakeLists.txt

@@ -289,7 +289,7 @@ if(Threads_FOUND)
         if(NOT result)
             # failed; under MinGW be sure to use the posix-threads and NOT the win32-threads version
             if(UPX_CONFIG_REQUIRE_THREADS OR (UPX_CONFIG_EXPECT_THREADS AND NOT UPX_CONFIG_DISABLE_THREADS))
-                message(WARNING "Threads FAILED ${f}: ${output}") # output from try_compile
+                message(WARNING "Threads FAILED ${f}: ${output}") # output from try_compile()
             endif()
             set(Threads_FOUND OFF)
             break()

misc/valgrind/musl.supp

@@ -9,6 +9,12 @@
    Memcheck:Cond
    fun:enframe
 }
+{
+   musl-malloc-get_nominal_size
+   Memcheck:Cond
+   fun:get_nominal_size
+   fun:__libc_free
+}
 
 # these are only needed for older musl versions
 

src/check/dt_xspan.cpp

@@ -101,9 +101,6 @@ TEST_CASE("basic xspan usage") {
         CHECK(b0.raw_size_in_bytes() == 0u);
         CHECK(bp.raw_size_in_bytes() == 0u);
 #endif
-        CHECK(raw_bytes(b0, 999999) == buf);
-        CHECK(raw_bytes(bp, 999999) == buf);
-
         CHECK(raw_bytes(c0, 4) == buf);
         CHECK(raw_index_bytes(c0, 1, 3) == buf + 1);
         CHECK(raw_bytes(cp, 4) == buf);
@@ -180,9 +177,6 @@ TEST_CASE("basic xspan usage") {
         CHECK(b0.raw_size_in_bytes() == 0u);
         CHECK(bp.raw_size_in_bytes() == 0u);
 #endif
-        CHECK(raw_bytes(b0, 999999) == buf);
-        CHECK(raw_bytes(bp, 999999) == buf);
-
         CHECK(raw_bytes(c0, 4) == buf);
         CHECK(raw_index_bytes(c0, 1, 3) == buf + 1);
         CHECK(raw_bytes(cp, 4) == buf);

src/stub/Makefile

@@ -167,7 +167,7 @@ all: $$(all.targets)
 	@echo "timestamp" > $@
 .DELETE_ON_ERROR: .upx-stubtools-stamp
 %/.tmp-stamp:
-	@mkdir -p $(dir $@)
+	@mkdir -p "$(dir $@)"
 	@echo "timestamp" > $@
 .PRECIOUS: %/.tmp-stamp
 .all-stamp: $$(STUBS)
@@ -241,15 +241,15 @@ define tc
 endef
 
 # default tools
-tc.default.bin2h      = $(PYTHON3) $(top_srcdir)/src/stub/scripts/bin2h.py --ident=auto-stub
+tc.default.bin2h      = $(PYTHON3) "$(top_srcdir)/src/stub/scripts/bin2h.py" --ident=auto-stub
 ##tc.default.bin2h-c    = $(call tc,bin2h) --compress=14,15,0
 tc.default.bin2h-c    = $(call tc,bin2h) --compress=0
-tc.default.brandelf   = $(PYTHON3) $(top_srcdir)/src/stub/scripts/brandelf.py $(if $(tc_bfdname),--bfdname=$(tc_bfdname))
+tc.default.brandelf   = $(PYTHON3) "$(top_srcdir)/src/stub/scripts/brandelf.py" $(if $(tc_bfdname),--bfdname=$(tc_bfdname))
-tc.default.gpp_inc    = $(PYTHON3) $(top_srcdir)/src/stub/scripts/gpp_inc.py
+tc.default.gpp_inc    = $(PYTHON3) "$(top_srcdir)/src/stub/scripts/gpp_inc.py"
-tc.default.gpp_mkdep  = $(PYTHON3) $(top_srcdir)/src/stub/scripts/gpp_inc.py -o /dev/null
+tc.default.gpp_mkdep  = $(PYTHON3) "$(top_srcdir)/src/stub/scripts/gpp_inc.py" -o /dev/null
 tc.default.pp-as      = i386-linux-gcc-3.4.6 -E -nostdinc -x assembler-with-cpp -Wall
 tc.default.sstrip     = sstrip-20060518
-tc.default.xstrip     = $(PYTHON3) $(top_srcdir)/src/stub/scripts/xstrip.py
+tc.default.xstrip     = $(PYTHON3) "$(top_srcdir)/src/stub/scripts/xstrip.py"
 
 # default multiarch-binutils
 tc.default.m-ar       = multiarch-ar-2.17
@@ -319,7 +319,7 @@ endef
 
 define tc.default.f-embed_objinfo_without_xstrip_keep_dot_text
 	chmod a-x $1
-	$(call tc,objcopy)          -R .data -R .bss $1
+	$(call tc,objcopy) -R .data -R .bss $1
 	$(call tc,objcopy) -R .comment -R .note -R .note.GNU-stack -R .reginfo $1
 	$(call tc,objcopy) --strip-unneeded --keep-symbol=_start --keep-symbol=upx_so_main $1
 #

src/util/util.cpp

@@ -260,11 +260,11 @@ const char *upx_getenv(const char *envvar) noexcept {
 }
 
 // random value from libc; quality is not important for UPX
-int upx_rand(void) noexcept {
+int upx_rand() noexcept {
     return ::rand(); // NOLINT(clang-analyzer-security.insecureAPI.rand)
 }
 
-void upx_rand_init(void) noexcept {
+void upx_rand_init() noexcept {
     unsigned seed = 0;
     seed ^= UPX_VERSION_HEX;
 #if (!HAVE_GETTIMEOFDAY || (ACC_OS_DOS32 && defined(__DJGPP__))) && !defined(__wasi__)
@@ -420,6 +420,22 @@ void upx_std_stable_sort(void *array, size_t n, upx_compare_func_t compare) {
 #endif
 }
 
+#if UPX_CONFIG_USE_STABLE_SORT
+// instantiate function templates for all element sizes we need; efficient
+// run-time, but code size bloat (about 4KiB code size for each function
+// with my current libstdc++); not really needed as libc qsort() is
+// good enough for our use cases
+template void upx_std_stable_sort<1>(void *, size_t, upx_compare_func_t);
+template void upx_std_stable_sort<2>(void *, size_t, upx_compare_func_t);
+template void upx_std_stable_sort<4>(void *, size_t, upx_compare_func_t);
+template void upx_std_stable_sort<5>(void *, size_t, upx_compare_func_t);
+template void upx_std_stable_sort<8>(void *, size_t, upx_compare_func_t);
+template void upx_std_stable_sort<16>(void *, size_t, upx_compare_func_t);
+template void upx_std_stable_sort<32>(void *, size_t, upx_compare_func_t);
+template void upx_std_stable_sort<56>(void *, size_t, upx_compare_func_t);
+template void upx_std_stable_sort<72>(void *, size_t, upx_compare_func_t);
+#endif // UPX_CONFIG_USE_STABLE_SORT
+
 TEST_CASE("upx_memswap") {
     auto check4 = [](int off1, int off2, int len, int a, int b, int c, int d) {
         byte p[4] = {0, 1, 2, 3};
@@ -473,22 +489,6 @@ TEST_CASE("upx_memswap") {
     }
 }
 
-#if UPX_CONFIG_USE_STABLE_SORT
-// instantiate function templates for all element sizes we need; efficient
-// run-time, but code size bloat (about 4KiB code size for each function
-// with my current libstdc++); not really needed as libc qsort() is
-// good enough for our use cases
-template void upx_std_stable_sort<1>(void *, size_t, upx_compare_func_t);
-template void upx_std_stable_sort<2>(void *, size_t, upx_compare_func_t);
-template void upx_std_stable_sort<4>(void *, size_t, upx_compare_func_t);
-template void upx_std_stable_sort<5>(void *, size_t, upx_compare_func_t);
-template void upx_std_stable_sort<8>(void *, size_t, upx_compare_func_t);
-template void upx_std_stable_sort<16>(void *, size_t, upx_compare_func_t);
-template void upx_std_stable_sort<32>(void *, size_t, upx_compare_func_t);
-template void upx_std_stable_sort<56>(void *, size_t, upx_compare_func_t);
-template void upx_std_stable_sort<72>(void *, size_t, upx_compare_func_t);
-#endif // UPX_CONFIG_USE_STABLE_SORT
-
 #if !defined(DOCTEST_CONFIG_DISABLE) && DEBUG
 #if __cplusplus >= 202002L // use C++20 std::next_permutation() to test all permutations
 namespace {

src/util/util.h

@@ -181,8 +181,8 @@ noinline const char *upx_getenv(const char *envvar) noexcept;
 
 noinline void upx_memswap(void *a, void *b, size_t bytes) noexcept;
 
-noinline void upx_rand_init(void) noexcept;
+noinline void upx_rand_init() noexcept;
-noinline int upx_rand(void) noexcept;
+noinline int upx_rand() noexcept;
 
 typedef int(__acc_cdecl_qsort *upx_compare_func_t)(const void *, const void *);
 typedef void (*upx_sort_func_t)(void *array, size_t n, size_t element_size, upx_compare_func_t);

src/util/xspan_impl.h

@@ -43,12 +43,12 @@
 XSPAN_NAMESPACE_BEGIN
 
 // HINT: set env-var "UPX_DEBUG_DOCTEST_DISABLE=1" for improved debugging experience
-noreturn void xspan_fail_nullptr(void) may_throw;
+noreturn void xspan_fail_nullptr() may_throw;
-noreturn void xspan_fail_nullbase(void) may_throw;
+noreturn void xspan_fail_nullbase() may_throw;
-noreturn void xspan_fail_not_same_base(void) may_throw;
+noreturn void xspan_fail_not_same_base() may_throw;
-noreturn void xspan_fail_range_nullptr(void) may_throw;
+noreturn void xspan_fail_range_nullptr() may_throw;
-noreturn void xspan_fail_range_nullbase(void) may_throw;
+noreturn void xspan_fail_range_nullbase() may_throw;
-noreturn void xspan_fail_range_range(void) may_throw;
+noreturn void xspan_fail_range_range() may_throw;
 void xspan_check_range(const void *ptr, const void *base, ptrdiff_t size_in_bytes) may_throw;
 
 // help constructor to distinguish between number of elements and bytes

src/util/xspan_impl_common.h

@@ -34,11 +34,6 @@ class CSelf {
 #endif
 
 public:
-    typedef T element_type;
-    typedef typename std::add_lvalue_reference<T>::type reference;
-    typedef typename std::add_pointer<T>::type pointer;
-    typedef size_t size_type;
-
     // befriend all
     template <class>
     friend struct PtrOrSpan;
@@ -47,6 +42,11 @@ public:
     template <class>
     friend struct Span;
 
+    typedef T element_type;
+    typedef typename std::add_lvalue_reference<T>::type reference;
+    typedef typename std::add_pointer<T>::type pointer;
+    typedef size_t size_type;
+
 #if XSPAN_CONFIG_ENABLE_IMPLICIT_CONVERSION
 public:
     operator pointer() const noexcept { return ptr; }
@@ -449,9 +449,10 @@ public: // raw access
         if (bytes > 0) {
             if __acc_cte (!configRequirePtr && ptr == nullptr)
                 xspan_fail_nullptr();
-            if __acc_cte (configRequireBase || base != nullptr) {
+            if __acc_cte (configRequireBase || base != nullptr)
                 xspan_check_range(ptr, base, size_in_bytes - bytes);
-            }
+            if very_unlikely (__acc_cte(VALGRIND_CHECK_MEM_IS_ADDRESSABLE(ptr, bytes) != 0))
+                throwCantPack("raw_bytes valgrind-check-mem");
         }
         return ptr;
     }

src/util/xspan_impl_ptr.h

@@ -201,6 +201,10 @@ private:
     static forceinline pointer check_deref(pointer p, ptrdiff_t n) noexcept { return p + n; }
     static forceinline pointer check_add(pointer p, ptrdiff_t n) noexcept { return p + n; }
 
+    // disable taking the address => force passing by reference
+    // [I'm not too sure about this design decision, but we can always allow it if needed]
+    Self *operator&() const XSPAN_DELETED_FUNCTION;
+
 public: // raw access
     pointer raw_ptr() const noexcept { return ptr; }
 
@@ -209,6 +213,8 @@ public: // raw access
         if (bytes > 0) {
             if very_unlikely (ptr == nullptr)
                 xspan_fail_nullptr();
+            if very_unlikely (__acc_cte(VALGRIND_CHECK_MEM_IS_ADDRESSABLE(ptr, bytes) != 0))
+                throwCantPack("raw_bytes valgrind-check-mem");
         }
         return ptr;
     }