Backports for 1.10.4 (#54416)

Backported PRs:
- [x] #54010 <!-- Overload `Base.literal_pow` for `AbstractQ` -->
- [x] #54143 <!-- Fix `make install` from tarballs -->
- [x] #54151 <!-- LinearAlgebra: Correct zero element in
`_generic_matvecmul!` for block adj/trans -->
- [x] #54233 <!-- set MAX_OS_WRITE on unix -->
- [x] #54251 <!-- fix typo in gc_mark_memory8 when chunking a large
array -->
- [x] #54363 <!-- typeintersect: fix another stack overflow caused by
circular constraints -->
- [x] #54497 <!-- Make TestLogger thread-safe (introduce a lock) -->
- [x] #53796 <!-- Add a missing doc -->
- [x] #54465 <!-- typeintersect: conservative typevar subtitution during
`finish_unionall` -->
- [x] #54514 <!-- typeintersect: followup cleanup for the nothrow path
of type instantiation -->

Need manual backport:
- [ ] #52505 <!-- fix alignment of emit_unbox_store copy -->
- [ ] #53373 <!-- fix sysimage-native-code=no option with pkgimages -->
- [ ] #53815 <!-- create phantom task for GC threads -->
- [ ] #53984 <!-- Profile: fix heap snapshot is valid char check -->
- [ ] #54276 <!-- Fix solve for complex `Hermitian` with non-vanishing
imaginary part on diagonal -->

Contains multiple commits, manual intervention needed:
- [ ] #52854 <!-- Change to streaming out the heap snapshot data -->
- [ ] #53218 <!-- Fix interpreter_exec.jl test -->
- [ ] #53833 <!-- Profile: make heap snapshots viewable in vscode viewer
-->
- [ ] #54303 <!-- LinearAlgebra: improve type-inference in
Symmetric/Hermitian matmul -->
- [ ] #52694 <!-- Reinstate similar for AbstractQ for backward
compatibility -->

Non-merged PRs with backport label:
- [ ] #54471 <!-- Actually setup jit targets when compiling
packageimages instead of targeting only one -->
- [ ] #53452 <!-- RFC: allow Tuple{Union{}}, returning Union{} -->
- [ ] #51479 <!-- prevent code loading from lookin in the versioned
environment when building Julia -->

Author: KristofferC

Make.inc

@@ -1365,7 +1365,7 @@ ifeq ($(OS), WINNT)
 HAVE_SSP := 1
 OSLIBS += -Wl,--export-all-symbols -Wl,--version-script=$(BUILDROOT)/src/julia.expmap \
 	$(NO_WHOLE_ARCHIVE) -lpsapi -lkernel32 -lws2_32 -liphlpapi -lwinmm -ldbghelp -luserenv -lsecur32 -latomic
-JLDFLAGS += -Wl,--stack,8388608
+JLDFLAGS += -Wl,--stack,8388608 --disable-auto-import --disable-runtime-pseudo-reloc
 ifeq ($(ARCH),i686)
 JLDFLAGS += -Wl,--large-address-aware
 endif

Makefile

@@ -61,6 +61,10 @@ $(foreach link,base $(JULIAHOME)/test,$(eval $(call symlink_target,$(link),$$(bu
 julia_flisp.boot.inc.phony: julia-deps
 	@$(MAKE) $(QUIET_MAKE) -C $(BUILDROOT)/src julia_flisp.boot.inc.phony
 
+# Build the HTML docs (skipped if already exists, notably in tarballs)
+$(BUILDROOT)/doc/_build/html/en/index.html: $(shell find $(BUILDROOT)/base $(BUILDROOT)/doc \( -path $(BUILDROOT)/doc/_build -o -path $(BUILDROOT)/doc/deps -o -name *_constants.jl -o -name *_h.jl -o -name version_git.jl \) -prune -o -type f -print)
+	@$(MAKE) docs
+
 julia-symlink: julia-cli-$(JULIA_BUILD_MODE)
 ifeq ($(OS),WINNT)
 	echo '@"%~dp0/'"$$(echo '$(call rel_path,$(BUILDROOT),$(JULIA_EXECUTABLE))')"'" %*' | tr / '\\' > $(BUILDROOT)/julia.bat
@@ -265,7 +269,7 @@ define stringreplace
 endef
 
 
-install: $(build_depsbindir)/stringreplace docs
+install: $(build_depsbindir)/stringreplace $(BUILDROOT)/doc/_build/html/en/index.html
 	@$(MAKE) $(QUIET_MAKE) $(JULIA_BUILD_MODE)
 	@for subdir in $(bindir) $(datarootdir)/julia/stdlib/$(VERSDIR) $(docdir) $(man1dir) $(includedir)/julia $(libdir) $(private_libdir) $(sysconfdir) $(private_libexecdir); do \
 		mkdir -p $(DESTDIR)$$subdir; \
@@ -523,7 +527,7 @@ app:
 darwinframework:
 	$(MAKE) -C $(JULIAHOME)/contrib/mac/framework
 
-light-source-dist.tmp: docs
+light-source-dist.tmp: $(BUILDROOT)/doc/_build/html/en/index.html
 ifneq ($(BUILDROOT),$(JULIAHOME))
 	$(error make light-source-dist does not work in out-of-tree builds)
 endif

base/linking.jl

@@ -110,7 +110,7 @@ function ld()
         # LLD supports mingw style linking
         flavor = "gnu"
         m = Sys.ARCH == :x86_64 ? "i386pep" : "i386pe"
-        default_args = `-m $m -Bdynamic --enable-auto-image-base --allow-multiple-definition`
+        default_args = `-m $m -Bdynamic --enable-auto-image-base --allow-multiple-definition --disable-auto-import --disable-runtime-pseudo-reloc`
     elseif Sys.isapple()
         flavor = "darwin"
         arch = Sys.ARCH == :aarch64 ? :arm64 : Sys.ARCH

base/stream.jl

@@ -122,7 +122,7 @@ const DEFAULT_READ_BUFFER_SZ = 10485760 # 10 MB
 if Sys.iswindows()
     const MAX_OS_WRITE = UInt(0x1FF0_0000) # 511 MB (determined semi-empirically, limited to 31 MB on XP)
 else
-    const MAX_OS_WRITE = UInt(typemax(Csize_t))
+    const MAX_OS_WRITE = UInt(0x7FFF_0000) # almost 2 GB (both macOS and linux have this kernel restriction, although only macOS documents it)
 end
 
 

cli/Makefile

@@ -17,8 +17,7 @@ LOADER_CFLAGS += -DGLIBCXX_LEAST_VERSION_SYMBOL=\"$(shell echo "$(CSL_NEXT_GLIBC
 endif
 
 ifeq ($(OS),WINNT)
-LOADER_LDFLAGS += -municode -mconsole -nostdlib --disable-auto-import \
-                  --disable-runtime-pseudo-reloc -lntdll -lkernel32 -lpsapi
+LOADER_LDFLAGS += -municode -mconsole -nostdlib -lntdll -lkernel32 -lpsapi
 else ifeq ($(OS),Linux)
 LOADER_LDFLAGS += -Wl,--no-as-needed -ldl -lpthread -rdynamic -lc -Wl,--as-needed
 else ifeq ($(OS),FreeBSD)

doc/src/base/math.md

@@ -70,6 +70,7 @@ Base.Math.tand
 Base.Math.sincosd
 Base.Math.sinpi
 Base.Math.cospi
+Base.Math.tanpi
 Base.Math.sincospi
 Base.sinh(::Number)
 Base.cosh(::Number)

src/aotcompile.cpp

@@ -1166,6 +1166,8 @@ static void materializePreserved(Module &M, Partition &partition) {
         GV.setInitializer(nullptr);
         GV.setLinkage(GlobalValue::ExternalLinkage);
         GV.setVisibility(GlobalValue::HiddenVisibility);
+        if (GV.getDLLStorageClass() != GlobalValue::DLLStorageClassTypes::DefaultStorageClass)
+            continue; // Don't mess with exported or imported globals
         GV.setDSOLocal(true);
     }
 
@@ -1645,6 +1647,7 @@ void jl_dump_native_impl(void *native_code,
             if (jl_small_typeof_copy) {
                 jl_small_typeof_copy->setVisibility(GlobalValue::HiddenVisibility);
                 jl_small_typeof_copy->setDSOLocal(true);
+                jl_small_typeof_copy->setDLLStorageClass(GlobalValue::DLLStorageClassTypes::DefaultStorageClass);
             }
         }
 
@@ -1673,16 +1676,17 @@ void jl_dump_native_impl(void *native_code,
         // reflect the address of the jl_RTLD_DEFAULT_handle variable
         // back to the caller, so that we can check for consistency issues
         GlobalValue *jlRTLD_DEFAULT_var = jl_emit_RTLD_DEFAULT_var(&metadataM);
-        addComdat(new GlobalVariable(metadataM,
-                                    jlRTLD_DEFAULT_var->getType(),
-                                    true,
-                                    GlobalVariable::ExternalLinkage,
-                                    jlRTLD_DEFAULT_var,
-                                    "jl_RTLD_DEFAULT_handle_pointer"), TheTriple);
-
         Type *T_size = DL.getIntPtrType(Context);
         Type *T_psize = T_size->getPointerTo();
 
+        auto FT = FunctionType::get(Type::getInt8Ty(Context)->getPointerTo()->getPointerTo(), {}, false);
+        auto F = Function::Create(FT, Function::ExternalLinkage, "get_jl_RTLD_DEFAULT_handle_addr", metadataM);
+        llvm::IRBuilder<> builder(BasicBlock::Create(Context, "top", F));
+        builder.CreateRet(jlRTLD_DEFAULT_var);
+        F->setLinkage(GlobalValue::ExternalLinkage);
+        if (TheTriple.isOSBinFormatCOFF())
+            F->setDLLStorageClass(GlobalValue::DLLStorageClassTypes::DLLExportStorageClass);
+
         if (TheTriple.isOSWindows()) {
             // Windows expect that the function `_DllMainStartup` is present in an dll.
             // Normal compilers use something like Zig's crtdll.c instead we provide a

src/builtins.c

@@ -1365,11 +1365,11 @@ JL_CALLABLE(jl_f_apply_type)
         jl_vararg_t *vm = (jl_vararg_t*)args[0];
         if (!vm->T) {
             JL_NARGS(apply_type, 2, 3);
-            return (jl_value_t*)jl_wrap_vararg(args[1], nargs == 3 ? args[2] : NULL, 1);
+            return (jl_value_t*)jl_wrap_vararg(args[1], nargs == 3 ? args[2] : NULL, 1, 0);
         }
         else if (!vm->N) {
             JL_NARGS(apply_type, 2, 2);
-            return (jl_value_t*)jl_wrap_vararg(vm->T, args[1], 1);
+            return (jl_value_t*)jl_wrap_vararg(vm->T, args[1], 1, 0);
         }
     }
     else if (jl_is_unionall(args[0])) {
@@ -2060,7 +2060,7 @@ void jl_init_primitives(void) JL_GC_DISABLED
     add_builtin("Tuple", (jl_value_t*)jl_anytuple_type);
     add_builtin("TypeofVararg", (jl_value_t*)jl_vararg_type);
     add_builtin("SimpleVector", (jl_value_t*)jl_simplevector_type);
-    add_builtin("Vararg", (jl_value_t*)jl_wrap_vararg(NULL, NULL, 0));
+    add_builtin("Vararg", (jl_value_t*)jl_wrap_vararg(NULL, NULL, 0, 0));
 
     add_builtin("Module", (jl_value_t*)jl_module_type);
     add_builtin("MethodTable", (jl_value_t*)jl_methtable_type);

src/codegen.cpp

@@ -493,9 +493,12 @@ struct JuliaVariable {
         if (GlobalValue *V = m->getNamedValue(name))
             return cast<GlobalVariable>(V);
         auto T_size = m->getDataLayout().getIntPtrType(m->getContext());
-        return new GlobalVariable(*m, _type(T_size),
+        auto var = new GlobalVariable(*m, _type(T_size),
                 isconst, GlobalVariable::ExternalLinkage,
                 NULL, name);
+        if (Triple(m->getTargetTriple()).isOSWindows())
+            var->setDLLStorageClass(GlobalValue::DLLStorageClassTypes::DLLImportStorageClass); // Cross-library imports must be explicit for COFF (Windows)
+        return var;
     }
     GlobalVariable *realize(jl_codectx_t &ctx);
 };
@@ -1786,9 +1789,6 @@ static inline GlobalVariable *prepare_global_in(Module *M, GlobalVariable *G)
                 G->isConstant(), GlobalVariable::ExternalLinkage,
                 nullptr, G->getName(), nullptr, G->getThreadLocalMode());
         proto->copyAttributesFrom(G);
-        // DLLImport only needs to be set for the shadow module
-        // it just gets annoying in the JIT
-        proto->setDLLStorageClass(GlobalValue::DefaultStorageClass);
         return proto;
     }
     return cast<GlobalVariable>(local);

src/gf.c

@@ -735,7 +735,7 @@ static jl_value_t *inst_varargp_in_env(jl_value_t *decl, jl_svec_t *sparams)
                 vm = T_has_tv ? jl_type_unionall(v, T) : T;
                 if (N_has_tv)
                     N = NULL;
-                vm = (jl_value_t*)jl_wrap_vararg(vm, N, 1); // this cannot throw for these inputs
+                vm = (jl_value_t*)jl_wrap_vararg(vm, N, 1, 0); // this cannot throw for these inputs
             }
             sp++;
             decl = ((jl_unionall_t*)decl)->body;
@@ -984,7 +984,7 @@ static void jl_compilation_sig(
             // avoid Vararg{Type{Type{...}}}
             if (jl_is_type_type(type_i) && jl_is_type_type(jl_tparam0(type_i)))
                 type_i = (jl_value_t*)jl_type_type;
-            type_i = (jl_value_t*)jl_wrap_vararg(type_i, (jl_value_t*)NULL, 1); // this cannot throw for these inputs
+            type_i = (jl_value_t*)jl_wrap_vararg(type_i, (jl_value_t*)NULL, 1, 0); // this cannot throw for these inputs
         }
         else {
             type_i = inst_varargp_in_env(decl, sparams);