Release Cosmopolitan v3.8.0

This change switches c++ exception handling from sjlj to standard dwarf. It's needed because clang for aarch64 doesn't support sjlj. It turns out that libunwind had a bare-metal configuration that made this easy to do. This change gets the new experimental cosmocc -mclang flag in a state of working so well that it can now be used to build all of llamafile and it goes 3x faster in terms of build latency, without trading away any perf. The int_fast16_t and int_fast32_t types are now always defined as 32-bit in the interest of having more abi consistency between cosmocc -mgcc and -mclang mode.
Delete superfluous function
2026-11-13 04:34:03 -10:00 · 2024-08-30 20:14:07 -07:00 · 2024-08-29 23:51:22 -07:00 · 2024-08-29 23:51:05 -07:00 · 2024-08-29 19:07:05 -07:00 · 2024-08-29 16:12:53 -07:00
2129 changed files with 214570 additions and 10295 deletions
@@ -27,3 +27,5 @@ da8baf2aa5ce93b958aca90a0ae69f537806324b
 369f9740de4534c28d0e81ab2afc99decbb9a3e6
 # Get rid of .internal.h convention in LIBC_INTRIN
 86d884cce24d773e298a2714c1e3d91ecab9be45
+# Remove .internal from more header filenames
+31194165d2afca36c2315a6e7ca2f0797dde09e3
@@ -0,0 +1,24 @@
+name: Nightly cosmocc
+on:
+  schedule:
+    # https://crontab.guru/#37_4_*_*_*
+    - cron: "37 4 * * *"
+  workflow_dispatch:
+concurrency:
+  group: ${{ github.workflow }}
+  cancel-in-progress: true
+jobs:
+  build-cosmocc:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - run: |
+          sudo cp build/bootstrap/ape.elf /usr/bin/ape
+          sudo sh -c "echo ':APE:M::MZqFpD::/usr/bin/ape:' >/proc/sys/fs/binfmt_misc/register"
+      - run: tool/cosmocc/package.sh
+      # https://github.com/actions/upload-artifact/issues/590
+      - uses: actions/[email protected]
+        with:
+          name: cosmocc
+          path: cosmocc
+          compression-level: 9
@@ -0,0 +1,36 @@
+{
+    "C_Cpp.default.compilerPath": ".cosmocc/3.8.0/bin/aarch64-linux-cosmo-c++",
+    "C_Cpp.default.compilerArgs": [
+        "-nostdinc",
+        "-nostdlib",
+        "-iquote.",
+        "-isystemlibc/isystem",
+        "-isystemthird_party/libcxx",
+        "-includelibc/integral/normalize.inc",
+        "-D_COSMO_SOURCE",
+        "-D__aarch64__"
+    ],
+    "[c]": {
+        "editor.tabSize": 2,
+        "editor.insertSpaces": true
+    },
+    "[cpp]": {
+        "editor.tabSize": 2,
+        "editor.insertSpaces": true
+    },
+    "[makefile]": {
+        "editor.tabSize": 8,
+        "editor.insertSpaces": false
+    },
+    "[make]": {
+        "editor.tabSize": 8,
+        "editor.insertSpaces": false
+    },
+    "[assembly]": {
+        "editor.tabSize": 8,
+        "editor.insertSpaces": true
+    },
+    "files.associations": {
+        "log.h": "c"
+    }
+}
@@ -115,10 +115,8 @@ ZIPCOPY = $(BOOTSTRAP)/zipcopy
 PECHECK = $(BOOTSTRAP)/pecheck
 FIXUPOBJ = $(BOOTSTRAP)/fixupobj
 OBJBINCOPY = $(BOOTSTRAP)/objbincopy
-MKDIR = build/bootstrap/mkdir -p
-COMPILE = build/bootstrap/compile -V9 -M2048m -P8192 $(QUOTA)
-
-IGNORE := $(shell $(MKDIR) $(TMPDIR))
+MKDIR = $(BOOTSTRAP)/mkdir.ape -p
+COMPILE = $(BOOTSTRAP)/compile.ape -V9 -M2048m -P8192 $(QUOTA)

 # the default build modes is empty string
 # on x86_64 hosts, MODE= is the same as MODE=x86_64
@@ -134,14 +132,13 @@ endif

 ifneq ($(findstring aarch64,$(MODE)),)
 ARCH = aarch64
-HOSTS ?= pi studio freebsdarm
+HOSTS ?= pi pi5 studio freebsdarm
 else
 ARCH = x86_64
 HOSTS ?= freebsd rhel7 xnu openbsd netbsd win10
 endif

 ZIPOBJ_FLAGS += -a$(ARCH)
-IGNORE := $(shell $(MKDIR) $(TMPDIR))

 export ADDR2LINE
 export LC_ALL
@@ -150,10 +147,12 @@ export MODE
 export SOURCE_DATE_EPOCH
 export TMPDIR

-COSMOCC = .cosmocc/3.6.0
+COSMOCC = .cosmocc/3.8.0
 BOOTSTRAP = $(COSMOCC)/bin
 TOOLCHAIN = $(COSMOCC)/bin/$(ARCH)-linux-cosmo-
-DOWNLOAD := $(shell build/download-cosmocc.sh $(COSMOCC) 3.6.0 4918c45ac3e0972ff260e2a249e25716881e39fb679d5e714ae216a2ef6c3f7e)
+DOWNLOAD := $(shell build/download-cosmocc.sh $(COSMOCC) 3.8.0 813c6b2f95062d2e0a845307a79505424cb98cb038e8013334f8a22e3b92a474)
+
+IGNORE := $(shell $(MKDIR) $(TMPDIR))

 AS = $(TOOLCHAIN)as
 CC = $(TOOLCHAIN)gcc
@@ -541,7 +540,7 @@ COSMOCC_HDRS =								\
 	$(foreach x,$(COSMOCC_PKGS),$($(x)_HDRS))			\
 	$(foreach x,$(COSMOCC_PKGS),$($(x)_INCS))

-o/cosmocc.h.txt: Makefile
+o/cosmocc.h.txt: Makefile libc $(MAKEFILES) $(call uniq,$(foreach x,$(HDRS) $(INCS),$(dir $(x)))) $(HDRS) $(INCS)
 	$(file >$@, $(call uniq,$(COSMOCC_HDRS)))

 COSMOPOLITAN_H_ROOT_HDRS =						\
@@ -45,10 +45,10 @@ o/$(MODE)/ape: $(APE)

 o/$(MODE)/ape/aarch64.lds:			\
 	ape/aarch64.lds				\
-	libc/zip.internal.h			\
+	libc/zip.h				\
 	libc/thread/tls.h			\
 	libc/calls/struct/timespec.h		\
-	libc/macros.internal.h			\
+	libc/macros.h				\
 	libc/str/str.h

 APE_LOADER_LDFLAGS =				\
@@ -162,8 +162,8 @@ o/$(MODE)/ape/ape-no-modify-self.o:		\
 		libc/dce.h			\
 		libc/elf/def.h			\
 		libc/thread/tls.h		\
-		libc/macho.internal.h		\
-		libc/macros.internal.h		\
+		libc/macho.h			\
+		libc/macros.h			\
 		libc/nexgen32e/uart.internal.h	\
 		libc/calls/metalfile.internal.h	\
 		libc/nt/pedef.internal.h	\
@@ -188,8 +188,8 @@ o/$(MODE)/ape/ape-copy-self.o:			\
 		libc/dce.h			\
 		libc/elf/def.h			\
 		libc/thread/tls.h		\
-		libc/macho.internal.h		\
-		libc/macros.internal.h		\
+		libc/macho.h			\
+		libc/macros.h			\
 		libc/nexgen32e/uart.internal.h	\
 		libc/calls/metalfile.internal.h	\
 		libc/nt/pedef.internal.h	\
@@ -246,8 +246,6 @@ o/$(MODE)/ape:	$(APE_CHECKS)			\
 		o/$(MODE)/ape/ape.lds		\
 		o/$(MODE)/ape/ape.elf		\
 		o/$(MODE)/ape/ape.macho		\
-		o/$(MODE)/ape/ape-copy-self.o	\
-		o/$(MODE)/ape/ape-no-modify-self.o

 endif

@@ -261,8 +259,8 @@ o/$(MODE)/ape/ape.o:				\
 		libc/thread/tls.h		\
 		ape/ape.internal.h		\
 		ape/macros.internal.h		\
-		libc/macho.internal.h		\
-		libc/macros.internal.h		\
+		libc/macho.h			\
+		libc/macros.h			\
 		libc/sysv/consts/prot.h		\
 		libc/nt/pedef.internal.h	\
 		libc/runtime/pc.internal.h	\
@@ -283,7 +281,7 @@ o/$(MODE)/ape/ape.lds:				\
 		libc/dce.h			\
 		libc/elf/def.h			\
 		libc/elf/pf2prot.internal.h	\
-		libc/macros.internal.h		\
+		libc/macros.h			\
 		libc/nt/pedef.internal.h	\
 		libc/str/str.h			\
-		libc/zip.internal.h
+		libc/zip.h
@@ -103,10 +103,8 @@ SECTIONS {
    *(.eh_frame_entry .eh_frame_entry.*)
  }

-  .eh_frame : ONLY_IF_RO {
-    KEEP(*(.eh_frame))
-    *(.eh_frame.*)
-  }
+  __eh_frame_hdr_start = SIZEOF(.eh_frame_hdr) > 0 ? ADDR(.eh_frame_hdr) : 0;
+  __eh_frame_hdr_end = SIZEOF(.eh_frame_hdr) > 0 ? . : 0;

  .gcc_except_table : ONLY_IF_RO {
    *(.gcc_except_table .gcc_except_table.*)
@@ -127,9 +125,11 @@ SECTIONS {
  . += CONSTANT(MAXPAGESIZE);
  . = DATA_SEGMENT_ALIGN(CONSTANT(MAXPAGESIZE), CONSTANT(COMMONPAGESIZE));

-  .eh_frame : ONLY_IF_RW {
+  .eh_frame : {
+    __eh_frame_start = .;
    KEEP(*(.eh_frame))
    *(.eh_frame.*)
+    __eh_frame_end = .;
  }

  .gnu_extab : ONLY_IF_RW {
@@ -16,6 +16,12 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
+#ifndef __APPLE__
+#error "ape/ape-m1.c is for apple silicon. chances you want ape/loader.c"
+#endif
+#ifndef __aarch64__
+#error "ape/ape-m1.c is for apple silicon; you want: make o//ape/ape.macho"
+#endif
 #include <assert.h>
 #include <dispatch/dispatch.h>
 #include <dlfcn.h>
@@ -37,7 +37,7 @@
 #include "libc/calls/metalfile.internal.h"
 #include "libc/dce.h"
 #include "libc/elf/def.h"
-#include "libc/macho.internal.h"
+#include "libc/macho.h"
 #include "libc/nexgen32e/uart.internal.h"
 #include "libc/nt/pedef.internal.h"
 #include "libc/runtime/pc.internal.h"
@@ -310,7 +310,7 @@ SECTIONS {
    . = ALIGN(__privileged_end > __privileged_start ? CONSTANT(COMMONPAGESIZE) : 0);
 /*END: morphable code */
    __privileged_start = .;
-    *(.privileged)
+    *(.privileged .privileged.*)
    __privileged_end = .;

    KEEP(*(.ape.pad.text))
@@ -329,6 +329,10 @@ SECTIONS {
    *(.ubsan.types)
    *(.ubsan.data)

+    __eh_frame_hdr_start_actual = .;
+    *(.eh_frame_hdr)
+    __eh_frame_hdr_end_actual = .;
+
    /* Legal Notices */
    __notices = .;
    KEEP(*(.notice))
@@ -422,6 +426,11 @@ SECTIONS {
    KEEP(*(.dtors))
    __fini_array_end = .;

+    __eh_frame_start = .;
+    KEEP(*(.eh_frame))
+    *(.eh_frame.*)
+    __eh_frame_end = .;
+
 /*BEGIN: Post-Initialization Read-Only */
    . = ALIGN(. != 0 ? __SIZEOF_POINTER__ : 0);
    KEEP(*(SORT_BY_NAME(.piro.relo.sort.*)))
@@ -601,6 +610,9 @@ ape_text_memsz = ape_text_filesz;
 ape_text_align = CONSTANT(COMMONPAGESIZE);
 ape_text_rva = RVA(ape_text_vaddr);

+__eh_frame_hdr_start = __eh_frame_hdr_end_actual > __eh_frame_hdr_start_actual ? __eh_frame_hdr_start_actual : 0;
+__eh_frame_hdr_end = __eh_frame_hdr_end_actual > __eh_frame_hdr_start_actual ? __eh_frame_hdr_end_actual : 0;
+
 /* we roundup here because xnu wants the file load segments page-aligned */
 /* but we don't want to add the nop padding to the ape program, so we'll */
 /* let ape.S dd read past the end of the file into the wrapping binaries */
@@ -16,7 +16,7 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 //	Calls _start() function of loaded program.
 //
@@ -16,10 +16,10 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/macho.internal.h"
+#include "libc/macho.h"
 #include "libc/sysv/consts/prot.h"
 #include "libc/dce.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 //	Apple Mach-O Executable Headers
 //	Fixups are applied by objbincopy
@@ -18,7 +18,7 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #ifndef APE_MACROS_H_
 #define APE_MACROS_H_
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #ifdef __ASSEMBLER__
 /* clang-format off */

@@ -18,7 +18,7 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/dce.h"
 #include "ape/ape.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 #ifdef __aarch64__

@@ -16,7 +16,7 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 //	Invokes system call.
 //
@@ -92,10 +92,7 @@ DEFAULT_COPTS ?=							\
 	-fno-gnu-unique							\
 	-fstrict-aliasing						\
 	-fstrict-overflow						\
-	-fno-semantic-interposition					\
-	-fno-dwarf2-cfi-asm						\
-	-fno-unwind-tables						\
-	-fno-asynchronous-unwind-tables
+	-fno-semantic-interposition

 ifeq ($(ARCH), x86_64)
 # Microsoft says "[a]ny memory below the stack beyond the red zone
@@ -115,14 +112,10 @@ ifeq ($(ARCH), aarch64)
 # - Cosmopolitan Libc uses x28 for thread-local storage because Apple
 #   forbids us from using tpidr_el0 too.
 #
-# - Cosmopolitan currently lacks an implementation of the runtime
-#   libraries needed by the -moutline-atomics flag
-#
 DEFAULT_COPTS +=							\
 	-ffixed-x18							\
 	-ffixed-x28							\
-	-fsigned-char							\
-	-mno-outline-atomics
+	-fsigned-char
 endif

 MATHEMATICAL =								\
@@ -143,8 +136,6 @@ DEFAULT_CFLAGS =							\

 DEFAULT_CXXFLAGS =							\
 	-std=gnu++23							\
-	-fno-rtti							\
-	-fno-exceptions							\
 	-fuse-cxa-atexit						\
 	-Wno-int-in-bool-context					\
 	-Wno-narrowing							\
@@ -6,14 +6,14 @@ if [ -n "$OBJDUMP" ]; then
 fi

 find_objdump() {
-  if [ -x .cosmocc/3.6.0/bin/$1-linux-cosmo-objdump ]; then
-    OBJDUMP=.cosmocc/3.6.0/bin/$1-linux-cosmo-objdump
-  elif [ -x .cosmocc/3.6.0/bin/$1-linux-musl-objdump ]; then
-    OBJDUMP=.cosmocc/3.6.0/bin/$1-linux-musl-objdump
-  elif [ -x "$COSMO/.cosmocc/3.6.0/bin/$1-linux-cosmo-objdump" ]; then
-    OBJDUMP="$COSMO/.cosmocc/3.6.0/bin/$1-linux-cosmo-objdump"
-  elif [ -x "$COSMO/.cosmocc/3.6.0/bin/$1-linux-musl-objdump" ]; then
-    OBJDUMP="$COSMO/.cosmocc/3.6.0/bin/$1-linux-musl-objdump"
+  if [ -x .cosmocc/3.8.0/bin/$1-linux-cosmo-objdump ]; then
+    OBJDUMP=.cosmocc/3.8.0/bin/$1-linux-cosmo-objdump
+  elif [ -x .cosmocc/3.8.0/bin/$1-linux-musl-objdump ]; then
+    OBJDUMP=.cosmocc/3.8.0/bin/$1-linux-musl-objdump
+  elif [ -x "$COSMO/.cosmocc/3.8.0/bin/$1-linux-cosmo-objdump" ]; then
+    OBJDUMP="$COSMO/.cosmocc/3.8.0/bin/$1-linux-cosmo-objdump"
+  elif [ -x "$COSMO/.cosmocc/3.8.0/bin/$1-linux-musl-objdump" ]; then
+    OBJDUMP="$COSMO/.cosmocc/3.8.0/bin/$1-linux-musl-objdump"
  else
    echo "error: toolchain not found (try running 'cosmocc --update' or 'make' in the cosmo monorepo)" >&2
    exit 1
@@ -4,5 +4,5 @@ UNAMES=$(uname -s)
 if [ x"$UNAMES" = x"Darwin" ] && [ x"$UNAMEM" = x"arm64" ]; then
  exec ape "$@"
 else
-  exec "$@"
+  exec rusage "$@"
 fi
@@ -241,8 +241,9 @@ class set
      private:
        friend class set;
        node_type* node_;
+        node_type* root_;

-        explicit reverse_iterator(node_type* node) : node_(node)
+        explicit reverse_iterator(node_type* node, node_type* root) : node_(node), root_(root)
        {
        }
    };
@@ -347,17 +348,17 @@ class set

    reverse_iterator rbegin()
    {
-        return reverse_iterator(rightmost(root_));
+        return reverse_iterator(rightmost(root_), root_);
    }

    const_reverse_iterator rbegin() const
    {
-        return const_reverse_iterator(rightmost(root_));
+        return const_reverse_iterator(rightmost(root_), root_);
    }

    const_reverse_iterator crbegin() const
    {
-        return const_reverse_iterator(rightmost(root_));
+        return const_reverse_iterator(rightmost(root_), root_);
    }

    iterator end() noexcept
@@ -377,17 +378,17 @@ class set

    reverse_iterator rend()
    {
-        return reverse_iterator(nullptr);
+        return reverse_iterator(nullptr, root_);
    }

    const_reverse_iterator rend() const
    {
-        return const_reverse_iterator(nullptr);
+        return const_reverse_iterator(nullptr, root_);
    }

    const_reverse_iterator crend() const
    {
-        return const_reverse_iterator(nullptr);
+        return const_reverse_iterator(nullptr, root_);
    }

    void clear() noexcept
@@ -1,6 +1,6 @@
 #ifndef COSMOPOLITAN_DSP_CORE_C161_H_
 #define COSMOPOLITAN_DSP_CORE_C161_H_
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 #define EXTRA_SHARP 2

@@ -1,7 +1,7 @@
 #ifndef COSMOPOLITAN_DSP_CORE_C161S_H_
 #define COSMOPOLITAN_DSP_CORE_C161S_H_
 #include "dsp/core/c161.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 __funline signed char C161S(signed char al, signed char bl, signed char cl) {
  short ax, bx, cx;
@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "dsp/core/core.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/math.h"
 #include "libc/mem/mem.h"

@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/limits.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/math.h"
 #include "libc/str/str.h"

@@ -20,7 +20,7 @@
 #include "libc/assert.h"
 #include "libc/dce.h"
 #include "libc/limits.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/math.h"
 #include "libc/str/str.h"

@@ -19,7 +19,7 @@
 #include "dsp/core/core.h"
 #include "dsp/core/q.h"
 #include "libc/dce.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/math.h"
 #include "libc/str/str.h"

@@ -1,6 +1,6 @@
 #ifndef COSMOPOLITAN_DSP_CORE_HALF_H_
 #define COSMOPOLITAN_DSP_CORE_HALF_H_
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 /**
 * Divides integer in half w/ rounding.
@@ -1,6 +1,6 @@
 #ifndef COSMOPOLITAN_DSP_CORE_KS8_H_
 #define COSMOPOLITAN_DSP_CORE_KS8_H_
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 /**
 * Performs 16-bit scaled rounded madd w/ eight coefficients or fewer.
@@ -1,7 +1,7 @@
 #ifndef COSMOPOLITAN_DSP_CORE_KSS8_H_
 #define COSMOPOLITAN_DSP_CORE_KSS8_H_
 #include "libc/limits.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 /**
 * Performs 16-bit scaled rounded saturated madd w/ eight coefficients or fewer.
@@ -1,7 +1,7 @@
 #ifndef COSMOPOLITAN_DSP_CORE_Q_H_
 #define COSMOPOLITAN_DSP_CORE_Q_H_
 #include "libc/limits.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/math.h"

 /**
@@ -18,7 +18,7 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "dsp/core/core.h"
 #include "libc/limits.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "third_party/aarch64/arm_neon.internal.h"
 #include "third_party/intel/emmintrin.internal.h"

@@ -16,7 +16,7 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 clamp4int256$core:
 	.leafprologue
@@ -35,7 +35,7 @@
 #include "libc/calls/struct/timespec.h"
 #include "libc/fmt/conv.h"
 #include "libc/log/log.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/math.h"
 #include "libc/mem/mem.h"
 #include "libc/str/str.h"
@@ -28,7 +28,7 @@
 │  SOFTWARE.                                                                   │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "dsp/mpeg/mpeg.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 __static_yoink("pl_mpeg_notice");

 /**
@@ -45,6 +45,12 @@ $(DSP_SCALE_A).pkg:				\
 		$(DSP_SCALE_A_OBJS)		\
 		$(foreach x,$(DSP_SCALE_A_DIRECTDEPS),$($(x)_A).pkg)

+ifeq ($(ARCH),x86_64)
+o/$(MODE)/dsp/scale/cdecimate2xuint8x8.o: private \
+		CFLAGS +=			\
+			-mssse3
+endif
+
 o/$(MODE)/dsp/scale/cdecimate2xuint8x8.o	\
 o/$(MODE)/dsp/scale/gyarados.o			\
 o/$(MODE)/dsp/scale/magikarp.o			\
@@ -25,7 +25,7 @@
 #include "libc/limits.h"
 #include "libc/log/check.h"
 #include "libc/log/log.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/math.h"
 #include "libc/mem/gc.h"
 #include "libc/mem/mem.h"
@@ -20,7 +20,7 @@
 #include "dsp/core/ks8.h"
 #include "dsp/core/kss8.h"
 #include "dsp/scale/cdecimate2xuint8x8.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nexgen32e/x86feature.h"
 #include "libc/str/str.h"
 #include "libc/x/x.h"
@@ -16,7 +16,7 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 //	TODO(jart): write me

@@ -21,7 +21,7 @@
 #include "libc/assert.h"
 #include "libc/limits.h"
 #include "libc/log/log.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/math.h"
 #include "libc/str/str.h"

@@ -18,7 +18,7 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "dsp/tty/quant.h"
 #include "libc/log/check.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 struct TtyRgb rgb2ttyi2f_(int r, int g, int b) {
  return rgb2ttyf((ttyrgb_m128){r, g, b} / 255);
@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "dsp/tty/quant.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 struct TtyRgb rgb2xterm24_(int r, int g, int b) {
  return (struct TtyRgb){MAX(MIN(r, 255), 0), MAX(MIN(g, 255), 0),
@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "dsp/tty/quant.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/math.h"
 #include "third_party/intel/xmmintrin.internal.h"

@@ -26,7 +26,7 @@
 #include "libc/limits.h"
 #include "libc/log/check.h"
 #include "libc/log/log.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/math.h"
 #include "libc/nexgen32e/x86feature.h"
 #include "libc/runtime/runtime.h"
@@ -24,7 +24,7 @@
 #include "libc/calls/termios.h"
 #include "libc/calls/ucontext.h"
 #include "libc/log/log.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/mem/gc.h"
 #include "libc/runtime/runtime.h"
 #include "libc/str/str.h"
@@ -16,7 +16,7 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 //	Returns index of minimum uint16 in array.
 //
@@ -16,7 +16,7 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 //	Returns index of minimum positive int16 in array.
 //
@@ -53,6 +53,7 @@ EXAMPLES_DIRECTDEPS =						\
 	LIBC_NEXGEN32E						\
 	LIBC_NT_ADVAPI32					\
 	LIBC_NT_IPHLPAPI					\
+	LIBC_NT_MEMORY						\
 	LIBC_NT_KERNEL32					\
 	LIBC_NT_NTDLL						\
 	LIBC_NT_USER32						\
@@ -93,6 +94,7 @@ EXAMPLES_DIRECTDEPS =						\
 	THIRD_PARTY_VQSORT					\
 	THIRD_PARTY_XED						\
 	THIRD_PARTY_LIBCXXABI					\
+	THIRD_PARTY_LIBUNWIND					\
 	THIRD_PARTY_ZLIB					\
 	TOOL_ARGS						\
 	TOOL_BUILD_LIB						\
@@ -0,0 +1,125 @@
+#if 0
+/*─────────────────────────────────────────────────────────────────╗
+│ To the extent possible under law, Justine Tunney has waived      │
+│ all copyright and related or neighboring rights to this file,    │
+│ as it is written in the following disclaimers:                   │
+│   • http://unlicense.org/                                        │
+│   • http://creativecommons.org/publicdomain/zero/1.0/            │
+╚─────────────────────────────────────────────────────────────────*/
+#endif
+#include <assert.h>
+#include <cosmo.h>
+#include <pthread.h>
+#include <stdatomic.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+// lockless push / pop tutorial
+//
+// this file demonstrates how to create a singly linked list that can be
+// pushed and popped across multiple threads, using only atomics. atomic
+// operations (rather than using a mutex) make push/pop go faster and it
+// ensures asynchronous signal safety too. therefore it will be safe for
+// use in a variety of contexts, such as signal handlers.
+
+#define THREADS    128
+#define ITERATIONS 10000
+
+// adjust mask based on alignment of list struct
+//
+// - 0x00fffffffffffff0 may be used if List* is always 16-byte aligned.
+//   We know that's the case here, because we call malloc() to create
+//   every List* object, and malloc() results are always max aligned.
+//
+// - 0x00fffffffffffff8 may be used if List* is always 8-byte aligned.
+//   This might be the case if you're pushing and popping stuff that was
+//   allocated from an array, to avoid malloc() calls. This has one
+//   fewer byte of safeguards against the ABA problem though.
+//
+// - 0x00fffffffffff000 may be used if List* is always page aligned.
+//   This is a good choice if you use mmap() to allocate each List*
+//   element, since it offers maximum protection against ABA.
+//
+// - only the highest byte of a 64-bit pointer is safe to use on our
+//   supported platforms. on most x86 and arm systems, it's possible to
+//   use the top sixteen bits. however that's not the case on more
+//   recent high end x86-64 systems that have pml5t.
+//
+#define MASQUE 0x00fffffffffffff0
+
+#define PTR(x)    ((uintptr_t)(x) & MASQUE)
+#define TAG(x)    ROL((uintptr_t)(x) & ~MASQUE, 8)
+#define ABA(p, t) ((uintptr_t)(p) | (ROR((uintptr_t)(t), 8) & ~MASQUE))
+#define ROL(x, n) (((x) << (n)) | ((x) >> (64 - (n))))
+#define ROR(x, n) (((x) >> (n)) | ((x) << (64 - (n))))
+
+struct List {
+  struct List* next;
+  int count;
+};
+
+atomic_uintptr_t list;
+
+void push(struct List* elem) {
+  uintptr_t tip;
+  assert(!TAG(elem));
+  for (tip = atomic_load_explicit(&list, memory_order_relaxed);;) {
+    elem->next = (struct List*)PTR(tip);
+    if (atomic_compare_exchange_weak_explicit(
+            &list, &tip, ABA(elem, TAG(tip) + 1), memory_order_release,
+            memory_order_relaxed))
+      break;
+    pthread_pause_np();
+  }
+}
+
+struct List* pop(void) {
+  uintptr_t tip;
+  struct List* elem;
+  tip = atomic_load_explicit(&list, memory_order_relaxed);
+  while ((elem = (struct List*)PTR(tip))) {
+    if (atomic_compare_exchange_weak_explicit(
+            &list, &tip, ABA(elem->next, TAG(tip) + 1), memory_order_acquire,
+            memory_order_relaxed))
+      break;
+    pthread_pause_np();
+  }
+  return elem;
+}
+
+void* tester(void* arg) {
+  struct List* elem;
+  for (int i = 0; i < ITERATIONS; ++i) {
+    while (!(elem = pop())) {
+      elem = malloc(sizeof(*elem));
+      elem->count = 0;
+      push(elem);
+    }
+    elem->count++;
+    push(elem);
+  }
+  return 0;
+}
+
+int main() {
+  printf("testing aba problem...");
+  fflush(stdout);
+  pthread_t th[THREADS];
+  for (int i = 0; i < THREADS; ++i)
+    pthread_create(&th[i], 0, tester, 0);
+  for (int i = 0; i < THREADS; ++i)
+    pthread_join(th[i], 0);
+  int sum = 0;
+  struct List* elem;
+  while ((elem = pop())) {
+    printf(" %d", elem->count);
+    sum += elem->count;
+    free(elem);
+  }
+  printf("\n");
+  assert(sum == ITERATIONS * THREADS);
+  printf("you are the dancing queen\n");
+  CheckForMemoryLeaks();
+}
@@ -10,7 +10,7 @@
 #include "libc/calls/calls.h"
 #include "libc/calls/struct/timespec.h"
 #include "libc/intrin/kprintf.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nt/enum/timezoneid.h"
 #include "libc/nt/struct/timezoneinformation.h"
 #include "libc/nt/time.h"
@@ -10,7 +10,7 @@
 #include "libc/calls/calls.h"
 #include "libc/fmt/conv.h"
 #include "libc/log/log.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/runtime/runtime.h"
 #include "libc/sock/sock.h"
 #include "libc/sock/struct/linger.h"
@@ -23,7 +23,7 @@
 #include "libc/inttypes.h"
 #include "libc/log/check.h"
 #include "libc/log/log.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/math.h"
 #include "libc/mem/arraylist2.internal.h"
 #include "libc/mem/mem.h"
@@ -47,7 +47,7 @@
 #include "libc/time.h"
 #include "libc/x/xasprintf.h"
 #include "libc/x/xsigaction.h"
-#include "libc/zip.internal.h"
+#include "libc/zip.h"
 #include "third_party/getopt/getopt.internal.h"
 #include "third_party/libcxx/vector"
 #include "tool/viz/lib/knobs.h"
@@ -0,0 +1,15 @@
+#if 0
+/*─────────────────────────────────────────────────────────────────╗
+│ To the extent possible under law, Justine Tunney has waived      │
+│ all copyright and related or neighboring rights to this file,    │
+│ as it is written in the following disclaimers:                   │
+│   • http://unlicense.org/                                        │
+│   • http://creativecommons.org/publicdomain/zero/1.0/            │
+╚─────────────────────────────────────────────────────────────────*/
+#endif
+#include <cosmo.h>
+#include <stdio.h>
+
+int main(int argc, char *argv[]) {
+  printf("%d\n", __get_cpu_count());
+}
@@ -1,4 +1,4 @@
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 //	Example assembly function.
 //
@@ -41,7 +41,7 @@
 #include "libc/fmt/conv.h"
 #include "libc/intrin/bswap.h"
 #include "libc/log/bsd.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/mem/mem.h"
 #include "libc/paths.h"
 #include "libc/runtime/runtime.h"
@@ -154,6 +154,66 @@ o/$(MODE)/libc/calls/sigcrashsig.o: private		\
 		CFLAGS +=				\
 			-Os

+# avoid legacy sse decoding penalty on avx systems
+o//libc/calls/cfmakeraw.o				\
+o//libc/calls/clock_gettime-xnu.o			\
+o//libc/calls/CPU_AND.o					\
+o//libc/calls/CPU_OR.o					\
+o//libc/calls/CPU_XOR.o					\
+o//libc/calls/dl_iterate_phdr.o				\
+o//libc/calls/dup-nt.o					\
+o//libc/calls/fcntl-nt.o				\
+o//libc/calls/flock-nt.o				\
+o//libc/calls/fstatfs-nt.o				\
+o//libc/calls/fstat-nt.o				\
+o//libc/calls/futimesat.o				\
+o//libc/calls/futimes.o					\
+o//libc/calls/getrlimit.o				\
+o//libc/calls/gettimeofday.o				\
+o//libc/calls/ioctl.o					\
+o//libc/calls/lutimes.o					\
+o//libc/calls/metaflock.o				\
+o//libc/calls/ntaccesscheck.o				\
+o//libc/calls/ntspawn.o					\
+o//libc/calls/open-nt.o					\
+o//libc/calls/pledge-linux.o				\
+o//libc/calls/ppoll.o					\
+o//libc/calls/preadv.o					\
+o//libc/calls/pselect.o					\
+o//libc/calls/pwritev.o					\
+o//libc/calls/read-nt.o					\
+o//libc/calls/readv.o					\
+o//libc/calls/readwrite-nt.o				\
+o//libc/calls/releasefd.o				\
+o//libc/calls/select.o					\
+o//libc/calls/sigaction.o				\
+o//libc/calls/sigenter-freebsd.o			\
+o//libc/calls/sigenter-netbsd.o				\
+o//libc/calls/sigenter-openbsd.o			\
+o//libc/calls/sigenter-xnu.o				\
+o//libc/calls/sigignore.o				\
+o//libc/calls/siginfo2cosmo.o				\
+o//libc/calls/signal.o					\
+o//libc/calls/sig.o					\
+o//libc/calls/sigtimedwait.o				\
+o//libc/calls/stat2cosmo.o				\
+o//libc/calls/statfs2cosmo.o				\
+o//libc/calls/statfs2statvfs.o				\
+o//libc/calls/tcgetattr-nt.o				\
+o//libc/calls/tcgetattr.o				\
+o//libc/calls/tcgetwinsize-nt.o				\
+o//libc/calls/tcsetattr-nt.o				\
+o//libc/calls/tcsetwinsize-nt.o				\
+o//libc/calls/termios2host.o				\
+o//libc/calls/timespec_sleep.o				\
+o//libc/calls/uname.o					\
+o//libc/calls/utimensat-old.o				\
+o//libc/calls/utimes.o					\
+o//libc/calls/winexec.o					\
+o//libc/calls/writev.o: private				\
+		COPTS +=				\
+			-mgeneral-regs-only
+
 # these assembly files are safe to build on aarch64
 o/$(MODE)/libc/calls/getcontext.o: libc/calls/getcontext.S
 	@$(COMPILE) -AOBJECTIFY.S $(OBJECTIFY.S) $(OUTPUT_OPTION) -c $<
@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/calls/struct/cpuset.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 void CPU_AND(cpu_set_t *d, cpu_set_t *x, cpu_set_t *y) {
  int i;
@@ -18,7 +18,7 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/calls/struct/cpuset.h"
 #include "libc/intrin/popcnt.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 int CPU_COUNT(cpu_set_t *set) {
  int i, c;
@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/calls/struct/cpuset.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 void CPU_OR(cpu_set_t *d, cpu_set_t *x, cpu_set_t *y) {
  int i;
@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/calls/struct/cpuset.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 void CPU_XOR(cpu_set_t *d, cpu_set_t *x, cpu_set_t *y) {
  int i;
@@ -18,7 +18,7 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/calls/syscall_support-nt.internal.h"
 #include "libc/errno.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nt/errors.h"
 #include "libc/nt/files.h"
 #include "libc/nt/process.h"
@@ -21,7 +21,7 @@
 #include "libc/calls/struct/timeval.h"
 #include "libc/calls/struct/timeval.internal.h"
 #include "libc/errno.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/sysv/consts/clock.h"
 #ifdef __x86_64__

@@ -17,13 +17,14 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/calls/internal.h"
-#include "libc/intrin/fds.h"
 #include "libc/calls/syscall-nt.internal.h"
 #include "libc/calls/syscall_support-nt.internal.h"
+#include "libc/intrin/fds.h"
 #include "libc/intrin/weaken.h"
 #include "libc/nt/enum/filetype.h"
 #include "libc/nt/files.h"
 #include "libc/nt/runtime.h"
+#include "libc/runtime/runtime.h"
 #include "libc/runtime/zipos.internal.h"
 #include "libc/sock/syscall_fd.internal.h"
 #include "libc/sysv/consts/o.h"
@@ -64,5 +65,7 @@ textwindows int sys_close_nt(int fd, int fildes) {
    default:
      break;
  }
+  if (f->cursor)
+    __cursor_unref(f->cursor);
  return CloseHandle(f->handle) ? 0 : __winerr();
 }
@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/calls/calls.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/runtime/runtime.h"

 /**
@@ -24,6 +24,7 @@
 #include "libc/calls/struct/sigset.internal.h"
 #include "libc/calls/syscall_support-nt.internal.h"
 #include "libc/errno.h"
+#include "libc/intrin/fds.h"
 #include "libc/intrin/kprintf.h"
 #include "libc/intrin/weaken.h"
 #include "libc/nt/files.h"
@@ -82,6 +83,7 @@ static textwindows int sys_dup_nt_impl(int oldfd, int newfd, int flags,

  g_fds.p[newfd] = g_fds.p[oldfd];
  g_fds.p[newfd].handle = handle;
+  __cursor_ref(g_fds.p[newfd].cursor);
  if (flags & _O_CLOEXEC) {
    g_fds.p[newfd].flags |= _O_CLOEXEC;
  } else {
@@ -20,18 +20,18 @@
 #include "libc/calls/calls.h"
 #include "libc/calls/createfileflags.internal.h"
 #include "libc/calls/internal.h"
-#include "libc/intrin/fds.h"
 #include "libc/calls/struct/flock.h"
 #include "libc/calls/struct/sigset.internal.h"
 #include "libc/calls/syscall-nt.internal.h"
 #include "libc/calls/syscall_support-nt.internal.h"
 #include "libc/calls/wincrash.internal.h"
 #include "libc/errno.h"
+#include "libc/intrin/fds.h"
 #include "libc/intrin/kprintf.h"
 #include "libc/intrin/weaken.h"
 #include "libc/limits.h"
 #include "libc/log/backtrace.internal.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/mem/leaks.h"
 #include "libc/mem/mem.h"
 #include "libc/nt/createfile.h"
@@ -151,7 +151,7 @@ static textwindows int sys_fcntl_nt_lock(struct Fd *f, int fd, int cmd,
    case SEEK_SET:
      break;
    case SEEK_CUR:
-      off = f->pointer + off;
+      off = f->cursor->shared->pointer + off;
      break;
    case SEEK_END: {
      int64_t size;
@@ -351,9 +351,14 @@ textwindows int sys_fcntl_nt(int fd, int cmd, uintptr_t arg) {
      }
      rc = 0;
    } else if (cmd == F_SETLK || cmd == F_SETLKW || cmd == F_GETLK) {
-      pthread_mutex_lock(&g_locks.mu);
-      rc = sys_fcntl_nt_lock(g_fds.p + fd, fd, cmd, arg);
-      pthread_mutex_unlock(&g_locks.mu);
+      struct Fd *f = g_fds.p + fd;
+      if (f->cursor) {
+        pthread_mutex_lock(&g_locks.mu);
+        rc = sys_fcntl_nt_lock(f, fd, cmd, arg);
+        pthread_mutex_unlock(&g_locks.mu);
+      } else {
+        rc = ebadf();
+      }
    } else if (cmd == F_DUPFD || cmd == F_DUPFD_CLOEXEC) {
      rc = sys_fcntl_nt_dupfd(fd, cmd, arg);
    } else {
@@ -27,6 +27,7 @@
 #include "libc/elf/tinyelf.internal.h"
 #include "libc/errno.h"
 #include "libc/intrin/directmap.h"
+#include "libc/intrin/promises.h"
 #include "libc/nt/memory.h"
 #include "libc/nt/runtime.h"
 #include "libc/runtime/runtime.h"
@@ -69,42 +70,35 @@ static int GetElfMachine(void) {
 }

 static bool IsMyDebugBinary(const char *path) {
+  void *addr;
  int64_t size;
  uintptr_t value;
  bool res = false;
  int fd, e = errno;
-  struct DirectMap dm;
-  BLOCK_CANCELATION;
  if ((fd = open(path, O_RDONLY | O_CLOEXEC, 0)) != -1) {
    // sanity test that this .com.dbg file (1) is an elf image, and (2)
    // contains the same number of bytes of code as our .com executable
    // which is currently running in memory.
    if ((size = lseek(fd, 0, SEEK_END)) != -1 &&
-        (dm = sys_mmap((void *)0x12345000000, size, PROT_READ, MAP_SHARED, fd,
-                       0))
-                .addr != MAP_FAILED) {
-      if (READ32LE((char *)dm.addr) == READ32LE("\177ELF") &&
-          ((Elf64_Ehdr *)dm.addr)->e_machine == GetElfMachine() &&
-          GetElfSymbolValue(dm.addr, "_etext", &value)) {
+        (addr = mmap(0, size, PROT_READ, MAP_SHARED, fd, 0)) != MAP_FAILED) {
+      if (READ32LE((char *)addr) == READ32LE("\177ELF") &&
+          ((Elf64_Ehdr *)addr)->e_machine == GetElfMachine() &&
+          GetElfSymbolValue(addr, "_etext", &value)) {
        res = !_etext || value == (uintptr_t)_etext;
      }
-      if (!IsWindows()) {
-        sys_munmap(dm.addr, size);
-      } else {
-        CloseHandle(dm.maphandle);
-        UnmapViewOfFile(dm.addr);
-      }
+      munmap(addr, size);
    }
    close(fd);
  }
-  ALLOW_CANCELATION;
  errno = e;
  return res;
 }

 static void FindDebugBinaryInit(void) {
  const char *comdbg;
-  if ((comdbg = getenv("COMDBG")) && IsMyDebugBinary(comdbg)) {
+  if (issetugid())
+    return;
+  if ((comdbg = getenv("COMDBG"))) {
    g_comdbg.res = comdbg;
    return;
  }
@@ -123,9 +117,19 @@ static void FindDebugBinaryInit(void) {
 /**
 * Returns path of binary with the debug information, or null.
 *
- * @return path to debug binary, or NULL
+ * You can specify the COMDBG environment variable, with the path of the
+ * debug binary, in case the automatic heuristics fail. What we look for
+ * is GetProgramExecutableName() with ".dbg", ".com.dbg", etc. appended.
+ *
+ * @return path to debug binary, or NULL if we couldn't find it
+ * @asyncsignalsafe
 */
 const char *FindDebugBinary(void) {
  cosmo_once(&g_comdbg.once, FindDebugBinaryInit);
  return g_comdbg.res;
 }
+
+// pay startup cost to make this signal safe from the user's perspective
+__attribute__((__constructor__(10))) static void FindDebugBinaryCtor(void) {
+  cosmo_once(&g_comdbg.once, FindDebugBinaryInit);
+}
@@ -27,7 +27,7 @@
 #include "libc/intrin/atomic.h"
 #include "libc/intrin/bsr.h"
 #include "libc/intrin/strace.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/mem/alloca.h"
 #include "libc/nt/enum/fileflagandattributes.h"
 #include "libc/nt/enum/fileinfobyhandleclass.h"
@@ -23,7 +23,7 @@
 #include "libc/calls/struct/statfs.internal.h"
 #include "libc/calls/syscall_support-nt.internal.h"
 #include "libc/limits.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nt/enum/fsinformationclass.h"
 #include "libc/nt/enum/status.h"
 #include "libc/nt/files.h"
@@ -16,7 +16,7 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 //	Gets machine state.
 //
@@ -30,39 +30,63 @@

 int sys_getcpu(unsigned *opt_cpu, unsigned *opt_node, void *tcache);

+/**
+ * Determines ID of CPU on which thread is currently scheduled.
+ *
+ * This is the same as sched_getcpu(), except it also supports returning
+ * the ID of the current NUMA node. On some platforms this functionality
+ * isn't available, in which case `out_opt_node` is always be set to 0.
+ */
 int getcpu(unsigned *out_opt_cpu, unsigned *out_opt_node) {
-  unsigned cpu;
-  unsigned node;
-  if (X86_HAVE(RDTSCP)) {
+
+  if (IsWindows()) {
+    struct NtProcessorNumber pn;
+    if (out_opt_cpu) {
+      GetCurrentProcessorNumberEx(&pn);
+      *out_opt_cpu = 64 * pn.Group + pn.Number;
+    }
+    if (out_opt_node) {
+      unsigned short node16;
+      if (GetNumaProcessorNodeEx(&pn, &node16)) {
+        *out_opt_node = node16;
+      } else {
+        return __winerr();
+      }
+    }
+    return 0;
+  }
+
+#ifdef __x86_64__
+  if (X86_HAVE(RDTSCP) && (IsLinux() || IsFreebsd())) {
    unsigned tsc_aux;
    rdtscp(&tsc_aux);
-    cpu = TSC_AUX_CORE(tsc_aux);
-    node = TSC_AUX_NODE(tsc_aux);
-  } else if (IsWindows()) {
-    struct NtProcessorNumber pn;
-    GetCurrentProcessorNumberEx(&pn);
-    cpu = 64 * pn.Group + pn.Number;
-    unsigned short node16;
-    if (GetNumaProcessorNodeEx(&pn, &node16)) {
-      node = node16;
-    } else {
-      return __winerr();
+    if (out_opt_cpu)
+      *out_opt_cpu = TSC_AUX_CORE(tsc_aux);
+    if (out_opt_node)
+      *out_opt_node = TSC_AUX_NODE(tsc_aux);
+    return 0;
+  }
+#endif
+
+  if (IsXnu() || IsOpenbsd() || IsNetbsd() || IsFreebsd()) {
+    if (out_opt_cpu) {
+      int rc = sched_getcpu();
+      if (rc == -1)
+        return -1;
+      *out_opt_cpu = rc;
    }
-  } else if (IsAarch64()) {
-    long tpidr_el0;
-    asm("mrs\t%0,tpidr_el0" : "=r"(tpidr_el0));
-    cpu = tpidr_el0 & 255;
-    node = 0;
-  } else {
-    int rc = sys_getcpu(&cpu, &node, 0);
-    if (rc == -1)
-      return -1;
+    if (out_opt_node)
+      *out_opt_node = 0;
+    return 0;
  }
-  if (out_opt_cpu) {
+
+  unsigned cpu, node;
+  int rc = sys_getcpu(&cpu, &node, 0);
+  if (rc == -1)
+    return -1;
+  if (out_opt_cpu)
    *out_opt_cpu = cpu;
-  }
-  if (out_opt_node) {
+  if (out_opt_node)
    *out_opt_node = node;
-  }
  return 0;
 }
@@ -20,7 +20,7 @@
 #include "libc/calls/syscall_support-sysv.internal.h"
 #include "libc/dce.h"
 #include "libc/intrin/strace.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nt/enum/computernameformat.h"
 #include "libc/str/str.h"
 #include "libc/sysv/errfuns.h"
@@ -18,7 +18,7 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/calls/struct/rlimit.h"
 #include "libc/limits.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/runtime/runtime.h"
 #include "libc/sysv/consts/rlimit.h"

@@ -18,7 +18,7 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/calls/calls.h"
 #include "libc/calls/syscall_support-nt.internal.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nt/enum/computernameformat.h"
 #include "libc/nt/systeminfo.h"
 #include "libc/str/str.h"
@@ -21,7 +21,7 @@
 #include "libc/calls/syscall_support-nt.internal.h"
 #include "libc/dce.h"
 #include "libc/fmt/conv.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nt/accounting.h"
 #include "libc/runtime/runtime.h"
 #include "libc/thread/thread.h"
@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/dce.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 //	Obtains WIN32 magic path, e.g. GetTempPathA.
 //
@@ -27,7 +27,7 @@
 #include "libc/intrin/getenv.h"
 #include "libc/intrin/strace.h"
 #include "libc/limits.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nt/runtime.h"
 #include "libc/runtime/runtime.h"
 #include "libc/serialize.h"
@@ -32,7 +32,7 @@
 #include "libc/intrin/asmflag.h"
 #include "libc/intrin/strace.h"
 #include "libc/intrin/weaken.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nexgen32e/kcpuids.h"
 #include "libc/nexgen32e/rdtsc.h"
 #include "libc/nexgen32e/vendor.internal.h"
@@ -18,7 +18,7 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/calls/internal.h"
 #include "libc/intrin/atomic.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nt/accounting.h"
 #include "libc/str/str.h"

@@ -25,7 +25,7 @@
 #include "libc/intrin/atomic.h"
 #include "libc/intrin/strace.h"
 #include "libc/limits.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/runtime/runtime.h"
 #include "libc/str/str.h"

@@ -5,9 +5,9 @@ COSMOPOLITAN_C_START_
 int sys_getgroups(int size, uint32_t list[]);
 int sys_setgroups(size_t size, const uint32_t list[]);

-const char *DescribeGidList(char[128], int, int, const uint32_t list[]);
+const char *_DescribeGidList(char[128], int, int, const uint32_t list[]);
 #define DescribeGidList(rc, length, gidlist) \
-  DescribeGidList(alloca(128), rc, length, gidlist)
+  _DescribeGidList(alloca(128), rc, length, gidlist)

 COSMOPOLITAN_C_END_
 #endif /* COSMOPOLITAN_LIBC_CALLS_GROUPS_INTERNAL_H_ */
@@ -4,7 +4,7 @@
 #include "libc/intrin/fds.h"
 #include "libc/calls/struct/sigval.h"
 #include "libc/dce.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/stdbool.h"

 #define kSigactionMinRva 8 /* >SIG_{ERR,DFL,IGN,...} */
@@ -27,7 +27,7 @@
 #include "libc/intrin/fds.h"
 #include "libc/intrin/strace.h"
 #include "libc/intrin/weaken.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/mem/alloca.h"
 #include "libc/mem/mem.h"
 #include "libc/nt/console.h"
@@ -25,6 +25,5 @@
 bool IsApeLoadable(char buf[8]) {
  return READ32LE(buf) == READ32LE("\177ELF") ||
         READ64LE(buf) == READ64LE("MZqFpD='") ||
-         READ64LE(buf) == READ64LE("jartsr='") ||
-         READ64LE(buf) == READ64LE("APEDBG='");
+         READ64LE(buf) == READ64LE("jartsr='");
 }
@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/dce.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 #define BYTES 64

@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/dce.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"

 #define BYTES 64

@@ -19,6 +19,7 @@
 #include "libc/calls/calls.h"
 #include "libc/calls/internal.h"
 #include "libc/calls/syscall_support-nt.internal.h"
+#include "libc/intrin/fds.h"
 #include "libc/nt/enum/filetype.h"
 #include "libc/nt/files.h"
 #include "libc/nt/struct/byhandlefileinformation.h"
@@ -31,7 +32,7 @@ static textwindows int64_t GetPosition(struct Fd *f, int whence) {
    case SEEK_SET:
      return 0;
    case SEEK_CUR:
-      return f->pointer;
+      return f->cursor->shared->pointer;
    case SEEK_END: {
      struct NtByHandleFileInformation wst;
      if (!GetFileInformationByHandle(f->handle, &wst)) {
@@ -67,11 +68,14 @@ textwindows int64_t sys_lseek_nt(int fd, int64_t offset, int whence) {
  } else if (__isfdkind(fd, kFdFile)) {
    struct Fd *f = g_fds.p + fd;
    int filetype = GetFileType(f->handle);
-    if (filetype != kNtFileTypePipe && filetype != kNtFileTypeChar) {
+    if (filetype != kNtFileTypePipe &&  //
+        filetype != kNtFileTypeChar &&  //
+        f->cursor->shared) {
      int64_t res;
-      if ((res = Seek(f, offset, whence)) != -1) {
-        f->pointer = res;
-      }
+      __cursor_lock(f->cursor);
+      if ((res = Seek(f, offset, whence)) != -1)
+        f->cursor->shared->pointer = res;
+      __cursor_unlock(f->cursor);
      return res;
    } else {
      return espipe();
@@ -32,7 +32,7 @@
 #include "libc/intrin/directmap.h"
 #include "libc/intrin/kprintf.h"
 #include "libc/intrin/weaken.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/mem/mem.h"
 #include "libc/runtime/pc.internal.h"
 #include "libc/runtime/runtime.h"
@@ -24,7 +24,7 @@
 │ ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR        │
 │ OTHER DEALINGS IN THE SOFTWARE.                                              │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/calls/metalfile.internal.h"

 	.init.start 102,_init_metalfile
@@ -20,7 +20,7 @@
 #include "libc/dce.h"
 #include "libc/intrin/kprintf.h"
 #include "libc/intrin/strace.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nt/systeminfo.h"
 #include "libc/str/str.h"
 #include "libc/sysv/consts/o.h"
@@ -55,6 +55,19 @@ textwindows size_t __normntpath(char16_t *p, size_t n) {
      // matched "/../" or "/..$"
      while (j && p[j - 1] == '\\')
        --j;
+      if (j && p[j - 1] == '.') {
+        // matched "." before
+        if (j >= 2 && p[j - 2] == '.' &&  //
+            (j == 2 || p[j - 3] == '\\')) {
+          // matched "^.." or "/.." before
+          p[++j] = '.';
+          ++j;
+          continue;
+        } else if (j == 1 || p[j - 2] == '\\') {
+          // matched "^." or "/." before
+          continue;
+        }
+      }
      while (j && p[j - 1] != '\\')
        --j;
    } else {
@@ -19,7 +19,7 @@
 #include "libc/calls/internal.h"
 #include "libc/calls/syscall_support-nt.internal.h"
 #include "libc/intrin/strace.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nt/enum/fileflagandattributes.h"
 #include "libc/nt/files.h"
 #include "libc/nt/thunk/msabi.h"
@@ -16,7 +16,7 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 .privileged

 __restore_rt_netbsd:
@@ -39,7 +39,7 @@
 #include "libc/nt/struct/startupinfo.h"
 #include "libc/nt/struct/startupinfoex.h"
 #include "libc/proc/ntspawn.h"
-#include "libc/stdalign.internal.h"
+#include "libc/stdalign.h"
 #include "libc/str/str.h"
 #include "libc/sysv/errfuns.h"
 #ifdef __x86_64__
@@ -24,7 +24,8 @@
 #include "libc/calls/syscall-nt.internal.h"
 #include "libc/calls/syscall_support-nt.internal.h"
 #include "libc/errno.h"
-#include "libc/macros.internal.h"
+#include "libc/intrin/fds.h"
+#include "libc/macros.h"
 #include "libc/nt/console.h"
 #include "libc/nt/createfile.h"
 #include "libc/nt/enum/accessmask.h"
@@ -138,6 +139,7 @@ static textwindows int sys_open_nt_file(int dirfd, const char *file,
  int64_t handle;
  if ((handle = sys_open_nt_impl(dirfd, file, flags, mode,
                                 kNtFileFlagOverlapped)) != -1) {
+    g_fds.p[fd].cursor = __cursor_new();
    g_fds.p[fd].handle = handle;
    g_fds.p[fd].kind = kFdFile;
    g_fds.p[fd].flags = flags;
@@ -170,15 +172,15 @@ static textwindows int sys_open_nt_no_handle(int fd, int flags, int mode,

 static textwindows int sys_open_nt_dup(int fd, int flags, int mode, int oldfd) {
  int64_t handle;
-  if (!__isfdopen(oldfd)) {
+  if (!__isfdopen(oldfd))
    return enoent();
-  }
  if (DuplicateHandle(GetCurrentProcess(), g_fds.p[oldfd].handle,
                      GetCurrentProcess(), &handle, 0, true,
                      kNtDuplicateSameAccess)) {
    g_fds.p[fd] = g_fds.p[oldfd];
    g_fds.p[fd].handle = handle;
    g_fds.p[fd].mode = mode;
+    __cursor_ref(g_fds.p[fd].cursor);
    if (!sys_fcntl_nt_setfl(fd, flags)) {
      return fd;
    } else {
@@ -23,7 +23,7 @@
 #include "libc/calls/metalfile.internal.h"
 #include "libc/intrin/directmap.h"
 #include "libc/intrin/weaken.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/mem/mem.h"
 #include "libc/runtime/pc.internal.h"
 #include "libc/runtime/runtime.h"
Author	SHA1	Message	Date
Justine Tunney	c9152b6f14	Release Cosmopolitan v3.8.0 This change switches c++ exception handling from sjlj to standard dwarf. It's needed because clang for aarch64 doesn't support sjlj. It turns out that libunwind had a bare-metal configuration that made this easy to do. This change gets the new experimental cosmocc -mclang flag in a state of working so well that it can now be used to build all of llamafile and it goes 3x faster in terms of build latency, without trading away any perf. The int_fast16_t and int_fast32_t types are now always defined as 32-bit in the interest of having more abi consistency between cosmocc -mgcc and -mclang mode.	2024-08-30 20:14:07 -07:00
Justine Tunney	5b9862907c	Delete superfluous function	2024-08-29 23:51:22 -07:00
Justine Tunney	c2420860e6	Fix --ftrace	2024-08-29 23:51:05 -07:00
Gabriel Ravier	6baf6cdb10	Fix vfprintf and derived functions badly handling +/` flag conflict (#1269 )	2024-08-29 19:07:05 -07:00
Gabriel Ravier	06a1193b4d	Make it so the test harness closes fds up to 100 (#1268 )	2024-08-29 16:12:53 -07:00
Justine Tunney	884d89235f	Harden against aba problem	2024-08-26 20:01:55 -07:00
Justine Tunney	610c951f71	Fix the build	2024-08-26 16:44:05 -07:00
Justine Tunney	12ecaf8650	Modernize ipv4.games server The server was originally written before I implemented support for POSIX thread cancelation. We now use standard pthreads APIs instead of talking directly to NSYNC. This means we're no longer using NSYNC notes, which aren't as good as the POSIX thread cancelation support I added to NSYNC which was only made possible by making NSYNC part of libc. I believe it will solve a crash we observed recently with ipv4.games, courtesy of the individual who goes by the hacker alias Lambro.	2024-08-26 16:05:23 -07:00
Justine Tunney	185e957696	Detect implicit function declarations This was suppressed recently and it's the worst possible idea when doing greenfield software development with C. I'm so sorry it slipped through. If the C standards committee was smart they would change the standard so that implicit int becomes implicit long. Then problems such as this will never occur and we could even use traditional C safely if we wanted too.	2024-08-26 15:34:08 -07:00
Justine Tunney	ebe1cbb1e3	Add crash proofing to ipv4.games server	2024-08-26 12:57:28 -07:00
Justine Tunney	e7b586e7f8	Add preliminary support for cosmocc -mclang C++ code compiles very slowly with cosmocc, possibly because we're using LLVM LIBCXX with GCC, and LLVM doesn't work as hard to make GCC go fast. Therefore, it should be possible, to ask cosmocc to favor Clang over GCC under the hood. On llamafile, my intention's to use this to make certain files, e.g. llama.cpp/common.cpp, go from taking 17 seconds to 5 seconds This new -mclang flag isn't ready for production yet since there's still the question of how to get Clang to generate SJLJ exception code. If you use this, then it's recommended you also pass -fno-exceptions. The tradeoff is we're adding a 121mb binary to the cosmocc distribution. There are no plans as of yet to fully migrate to Clang since GCC is very good and has always treated us well.	2024-08-26 12:33:35 -07:00
Justine Tunney	111ec9a989	Fix bug we added to *NSYNC a while ago This is believed to fix a crash, that's possible in nsync_waiter_free_() when you call pthread_cond_timedwait(), or nsync_cv_wait_with_deadline() where an assertion can fail. Thanks ipv4.games for helping me find this!	2024-08-26 12:25:50 -07:00
Justine Tunney	f3ce684aef	Fix getpeername() bug on Windows The WIN32 getpeername() function returns ENOTCONN when it uses connect() the SOCK_NONBLOCK way. So we simply store the address, provided earlier.	2024-08-25 11:28:53 -07:00
Justine Tunney	908b7a82ca	Add VSCode settings	2024-08-25 11:02:31 -07:00
Justine Tunney	bb06230f1e	Avoid linker conflicts on DescribeFoo symbols These symbols belong to the user. It caused a confusing error for Blink.	2024-08-24 18:10:22 -07:00
Justine Tunney	38cc4b3c68	Get rid of some legacy code	2024-08-24 17:53:30 -07:00
Justine Tunney	37ca1badaf	Make Emacs load 2x faster	2024-08-23 20:08:05 -07:00
Justine Tunney	1a9f82bc9f	Romanize Hindi, Yiddish, Arabic, Cyrillic, etc.	2024-08-20 16:27:16 -07:00
Justine Tunney	df1aee7ce5	Upgrade superconfigure and monorepo toolchain See #1260	2024-08-20 08:46:21 -07:00
Justine Tunney	2d44142444	Get Meson builds working See #917	2024-08-19 08:40:18 -07:00
Justine Tunney	4bbc16e2cc	Add helpful error messages	2024-08-19 07:28:49 -07:00
Justine Tunney	863c704684	Add string similarity function	2024-08-17 16:45:07 -07:00
Justine Tunney	60e697f7b2	Move LoadZipArgs() to cosmo.h	2024-08-17 12:06:27 -07:00
Justine Tunney	4389f4709a	Expose wmempcpy() to _GNU_SOURCE	2024-08-17 08:13:22 -07:00
Justine Tunney	ca2c30c977	Release redbean v3.0.0	2024-08-17 06:45:35 -07:00
Justine Tunney	eb6e96f036	Change InfoZIP to not auto-append .zip to pathname	2024-08-17 06:45:23 -07:00
Justine Tunney	77be460290	Make Windows REPLs great again	2024-08-17 06:32:10 -07:00
Justine Tunney	8e14b27749	Make fread() more consistent with glibc	2024-08-17 02:57:22 -07:00
Justine Tunney	1d532ba3f8	Disable some anti-Musl Lua tests	2024-08-17 02:20:08 -07:00
Justine Tunney	b2a1811c01	Add missing pragma	2024-08-16 21:49:28 -07:00
Justine Tunney	2eda50929b	Add stdfloat header Fixes #1260	2024-08-16 21:38:00 -07:00
Justine Tunney	098638cc6c	Fix pthread_kill_test flake on qemu	2024-08-16 21:18:26 -07:00
Justine Tunney	732554ce3a	Release Cosmopolitan v3.7.1	2024-08-16 11:56:47 -07:00
Gavin Hayes	914d521090	Fix relative Windows path normalization (#1261 ) Fixes #1223	2024-08-16 11:55:49 -07:00
Justine Tunney	11d9fb521d	Make atomics faster on aarch64 This change implements the compiler runtime for ARM v8.1 ISE atomics and gets rid of the mandatory -mno-outline-atomics flag. It can dramatically speed things up, on newer ARM CPUs, as indicated by the changed lines in test/libc/thread/footek_test.c. In llamafile dispatching on hwcap atomic also shaved microseconds off synchronization barriers.	2024-08-16 11:14:46 -07:00
Justine Tunney	de0cde8def	Release Cosmopolitan v3.7.0	2024-08-16 07:43:10 -07:00
Justine Tunney	5bd22aef12	Experiment with supporting Windows Arm64 natively So far I haven't found any way to run native Arm64 code on Windows Arm64 without using MSVC. When I build a PE binary from scratch that should be a valid Windows Arm64 program, the OS refuses to run it. Possibly due to requiring additional content like XML manifests or relocation or control flow integrity data that isn't normally required on x64. I've also tried using VirtualAlloc2() to JIT an Arm64 native function, but VirtualAlloc2 always fails with invalid parameter. I tried using MSVC to create an ARM DLL that my x64 emulated program can link at runtime, to pass a function pointer with ARM code, but LoadLibrary() rejects ARM DLLs as invalid exe The only option left, is likely to write a new program like ape/ape-m1.c which can be compiled by MSVC to load and run an AARCH64 ELF executable. The emulated x64 binary would detect emulation using IsWow64Process2 and then drop the loader executable in a temporary folder, and re-launch the original executable, using the Arm64 segments of the cosmocc fat binary.	2024-08-16 06:43:59 -07:00
Justine Tunney	1671283f1a	Avoid clobbering errno	2024-08-15 23:54:14 -07:00
Justine Tunney	0a79c6961f	Make malloc scalable on all platforms It turns out sched_getcpu() didn't work on many platforms. So the system call now has tests and is well documented. We now employ new workarounds on platforms where it isn't supported in our malloc() implementation. It was previously the case that malloc() was only scalable on Linux/Windows for x86-64. Now the other platforms are scalable too.	2024-08-15 23:32:53 -07:00
Justine Tunney	3fd275f59f	Import optimized routines changes to exp10	2024-08-15 18:37:33 -07:00
Justine Tunney	2045e87b7c	Fix build issues	2024-08-15 18:37:33 -07:00
Jacob Hummer	24666e121d	add nightly cosmocc (artifact) builds (#1254 ) Using https://nightly.link/ with GitHub actions artifacts you can have a nightly build (but not a _release_ -- there's no releasing or pre-releasing happening) of cosmocc. Example URL if this PR were merged: https://nightly.link/jart/cosmopolitan/workflows/nightly-cosmocc/master/cosmocc.zip Or you can just download it directly from the GitHub "Actions" https://github.com/jart/cosmopolitan/actions workflow summary page of a particular run example from my own fork: ![image](https://github.com/user-attachments/assets/8ba708dd-8289-4f8b-932c-cf535ee86f62) could download by clicking on the artifact or by using third-party service to provide a link for unauthenticated requests (like wget or curl) https://nightly.link/jcbhmr/cosmopolitan/workflows/tool-cosmocc-package-sh/master/cosmocc.zip this would be useful for users who don't want to or can't figure out how to build cosmocc themselves (like Windows) but still want to use a nightly build since a fix hasn't been released as a release version yet. this would also be a good way to test the release process but instead of pushing the `cosmocc.zip` to _wherever it goes now_ you publish it as a github actions artifact for the very few nightly bleeding edge users to use & test. you don't have to use https://nightly.link or recommend it or anything; i just know its a cool way to wget or curl the URLs instead of downloading it via your browser web UI. particularly useful for remote/ssh/web-ide development.	2024-08-15 00:47:40 -04:00
Justine Tunney	ff1a0d1c40	Upgrade to superconfigure z0.0.51	2024-08-04 14:59:53 -07:00
Justine Tunney	7f0db3e3b9	Add last commit to .git-blame-ignore-revs	2024-08-04 12:54:30 -07:00
Justine Tunney	31194165d2	Remove .internal from more header filenames	2024-08-04 12:52:25 -07:00
Justine Tunney	c265c17d54	Fix the build	2024-08-04 07:08:28 -07:00
Justine Tunney	7499367060	Ignore -Wimplicit-function-declaration in cosmocc	2024-08-03 21:36:36 -07:00
Justine Tunney	3f26dfbb31	Share file offset across execve() on Windows This is a breaking change. It defines the new environment variable named _COSMO_FDS_V2 which is used for inheriting non-stdio file descriptors on execve() or posix_spawn(). No effort has been spent thus far integrating with the older variable. If a new binary launches the older ones or vice versa they'll only be able to pass stdin / stdout / stderr to each other therefore it's important that you upgrade all your cosmo binaries if you depend on this functionality. You'll be glad you did because inheritance of file descriptors is more aligned with the POSIX standard than before.	2024-08-03 17:48:00 -07:00
Justine Tunney	761c6ad615	Share file offset across processes This change ensures that if a file descriptor for an open disk file gets shared by multiple processes within a process tree, then lseek() changes will be visible across processes, and read() / write() are synchronized. Note this only applies to Windows, because UNIX kernels already do this.	2024-08-03 01:39:11 -07:00
Justine Tunney	a80ab3f8fe	Implement bf16 compiler runtime library	2024-08-02 02:04:53 -07:00
Gautham	9ebacb7892	Convert GCC 14 errors back to warnings (#1247 ) https://gcc.gnu.org/gcc-14/porting_to.html#warnings-as-errors Changing these to warnings helps build code with `cosmocc`. Perhaps this can be a patch to `cosmocc` or skipped entirely.	2024-08-01 21:42:40 -07:00
Justine Tunney	6ac3d3b804	Add precompiled header support to cosmocc	2024-07-31 06:04:41 -07:00
Justine Tunney	f8cfc89eba	Allow -c to be specified with -E in cosmocc	2024-07-31 02:09:15 -07:00
Justine Tunney	4ed4a1095a	Improve build latency	2024-07-31 01:21:27 -07:00
Justine Tunney	8d8aecb6d9	Avoid legacy instruction penalties on x86	2024-07-31 01:02:38 -07:00
Justine Tunney	1fba310e22	Delete mislocated headers	2024-07-31 01:01:00 -07:00
Justine Tunney	bb815eafaf	Update Musl Libc code We now have implement all of Musl's localization code, the same way that Musl implements localization. You may need setlocale(LC_ALL, "C.UTF-8"), just in case anything stops working as expected.	2024-07-30 22:51:29 -07:00
Justine Tunney	d0360bf4bd	Introduce cosmoranlib	2024-07-29 19:18:52 -07:00
Justine Tunney	8cdb3e136b	Check in ruler summation experiments	2024-07-29 18:02:16 -07:00
Justine Tunney	3dab207351	Remove mkfifo() prototype	2024-07-29 07:42:54 -07:00
Justine Tunney	1092d9b7e8	Remove old usages of %n	2024-07-29 01:39:36 -07:00
Justine Tunney	d40acc60b1	Detect more x86 features	2024-07-29 00:16:29 -07:00
Justine Tunney	cf1559c448	Remove __threaded variable	2024-07-28 23:43:30 -07:00
Justine Tunney	01b09bc817	Support printf %n directive	2024-07-28 22:27:06 -07:00
Justine Tunney	c1a0b017e9	Fix the build	2024-07-28 21:02:04 -07:00
Justine Tunney	77d3a07ff2	Fix std::filesystem This change makes a second pass, at fixing the errno issue with libcxx's filesystem code. Previously, 89.01% of LLVM's test suite was passing and now 98.59% of their tests pass. Best of all, it's now possible for Clang to be built as a working APE binary that can to compile the Cosmopolitan repository. Please note it has only been vetted so far for some objects, and more work would obviously need to be done in cosmo, to fix warnings.	2024-07-28 17:31:21 -07:00
Justine Tunney	0d7c272d3f	Don't use sendfile() in libcxx	2024-07-28 17:31:21 -07:00
Justine Tunney	18964e5d76	Fix remove() directory on Windows	2024-07-28 17:31:21 -07:00
Justine Tunney	e18fe1e112	Freshen build/bootstrap/cocmd See https://news.ycombinator.com/item?id=41055121	2024-07-27 23:22:11 -07:00