Release Cosmopolitan v3.7.0

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.

.git-blame-ignore-revs

@@ -27,3 +27,5 @@ da8baf2aa5ce93b958aca90a0ae69f537806324b
 369f9740de4534c28d0e81ab2afc99decbb9a3e6
 # Get rid of .internal.h convention in LIBC_INTRIN
 86d884cce24d773e298a2714c1e3d91ecab9be45
+# Remove .internal from more header filenames
+31194165d2afca36c2315a6e7ca2f0797dde09e3

.github/workflows/nightly-cosmocc.yml

@@ -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

Makefile

@@ -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
@@ -141,7 +139,6 @@ 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/3.6.2
 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) 3.6.2 268aa82d9bfd774f76951b250f87b8edcefd5c754b8b409e1639641e8bd8d5bc)
+
+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 =						\

ape/BUILD.mk

@@ -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

ape/ape.S

@@ -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"

ape/launch.S

@@ -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.
 //

ape/loader-macho.S

@@ -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

ape/macros.internal.h

@@ -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 */
 

ape/start.S

@@ -18,7 +18,7 @@
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/dce.h"
 #include "ape/ape.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 
 #ifdef __aarch64__
 

ape/systemcall.S

@@ -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.
 //

dsp/core/c161.h

@@ -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
 

dsp/core/c161s.h

@@ -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;

dsp/core/double2byte.c

@@ -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"
 

dsp/core/float2short.c

@@ -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"
 

dsp/core/getintegercoefficients.c

@@ -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"
 

dsp/core/getintegercoefficients8.c

@@ -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"
 

dsp/core/half.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.

dsp/core/ks8.h

@@ -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.

dsp/core/kss8.h

@@ -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.

dsp/core/q.h

@@ -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"
 
 /**

dsp/core/sad16x8n.c

@@ -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"
 

dsp/mpeg/clamp4int256-core.S

@@ -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

dsp/mpeg/mpeg1.c

@@ -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"

dsp/mpeg/slowrgb.c

@@ -28,7 +28,7 @@
 │  SOFTWARE.                                                                   │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "dsp/mpeg/mpeg.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 __static_yoink("pl_mpeg_notice");
 
 /**

dsp/scale/gyarados.c

@@ -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"

dsp/scale/magikarp.c

@@ -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"

dsp/tty/mpsadbw.S

@@ -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
 

dsp/tty/rgb2ansi.c

@@ -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"
 

dsp/tty/rgb2ttyi2f.c

@@ -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);

dsp/tty/rgb2xterm24.c

@@ -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),

dsp/tty/rgb2xterm24f.c

@@ -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"
 

dsp/tty/ttyraster.c

@@ -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"

dsp/tty/ttyraw.c

@@ -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"

dsp/tty/windex-avx2.S

@@ -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.
 //

dsp/tty/windex-sse4.S

@@ -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.
 //

examples/BUILD.mk

@@ -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						\

examples/date.c

@@ -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"

examples/nc.c

@@ -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"

examples/nesemu1.cc

@@ -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"

examples/nproc.c

@@ -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());
+}

examples/package/lib/myasm.S

@@ -1,4 +1,4 @@
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 
 //	Example assembly function.
 //

examples/script.c

@@ -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"

libc/calls/BUILD.mk

@@ -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 $<

libc/calls/CPU_AND.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;

libc/calls/CPU_COUNT.c

@@ -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;

libc/calls/CPU_OR.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;

libc/calls/CPU_XOR.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_XOR(cpu_set_t *d, cpu_set_t *x, cpu_set_t *y) {
   int i;

libc/calls/chdir-nt.c

@@ -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"

libc/calls/clock_gettime-xnu.c

@@ -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__
 

libc/calls/close-nt.c

@@ -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();
 }

libc/calls/copy.c

@@ -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"
 
 /**

libc/calls/dup-nt.c

@@ -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 {

libc/calls/fcntl-nt.c

@@ -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 {

libc/calls/finddebugbinary.c

@@ -104,6 +104,8 @@ static bool IsMyDebugBinary(const char *path) {
 
 static void FindDebugBinaryInit(void) {
   const char *comdbg;
+  if (issetugid())
+    return;
   if ((comdbg = getenv("COMDBG")) && IsMyDebugBinary(comdbg)) {
     g_comdbg.res = comdbg;
     return;

libc/calls/fstat-nt.c

@@ -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"

libc/calls/fstatfs-nt.c

@@ -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"

libc/calls/getcontext.S

@@ -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.
 //

libc/calls/getcpu.c

@@ -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;
 }

libc/calls/getdomainname.c

@@ -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"

libc/calls/getdtablesize.c

@@ -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"
 

libc/calls/gethostname-nt.c

@@ -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"

libc/calls/getloadavg-nt.c

@@ -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"

libc/calls/getntsyspath.S

@@ -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.
 //

libc/calls/getprogramexecutablename.greg.c

@@ -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"

libc/calls/getrandom.c

@@ -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"

libc/calls/getuid-nt.c

@@ -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"
 

libc/calls/getuid.c

@@ -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"
 

libc/calls/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,...} */

libc/calls/ioctl.c

@@ -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"

libc/calls/isapemagic.c

@@ -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='");
 }

libc/calls/kntsystemdirectory.S

@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/dce.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 
 #define BYTES 64
 

libc/calls/kntwindowsdirectory.S

@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/dce.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 
 #define BYTES 64
 

libc/calls/lseek-nt.c

@@ -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();

libc/calls/metalfile.c

@@ -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"

libc/calls/metalfile_init.S

@@ -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

libc/calls/mkntpath.c

@@ -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"

libc/calls/mkntpathat.c

@@ -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"

libc/calls/netbsdtramp.S

@@ -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:

libc/calls/ntspawn.c

@@ -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__

libc/calls/open-nt.c

@@ -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 {

libc/calls/openat-metal.c

@@ -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"

libc/calls/parsepromises.c

@@ -17,7 +17,7 @@
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
 #include "libc/calls/pledge.internal.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/str/str.h"
 
 static int FindPromise(const char *name) {

libc/calls/pledge-linux.c

@@ -29,7 +29,7 @@
 #include "libc/intrin/bsr.h"
 #include "libc/intrin/likely.h"
 #include "libc/intrin/promises.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/runtime/runtime.h"
 #include "libc/runtime/stack.h"
 #include "libc/sysv/consts/audit.h"

libc/calls/pledge.c

@@ -169,8 +169,7 @@
  *   turn APE binaries into static native binaries.
  *
  * - "prot_exec" allows mmap(PROT_EXEC) and mprotect(PROT_EXEC). This is
- *   needed to (1) code morph mutexes in __enable_threads(), and it's
- *   needed to (2) launch non-static or non-native executables, e.g.
+ *   needed to launch non-static or non-native executables, e.g.
  *   non-assimilated APE binaries, or dynamic-linked executables.
  *
  * - "unveil" allows unveil() to be called, as well as the underlying

libc/calls/poll-nt.c

@@ -28,7 +28,7 @@
 #include "libc/errno.h"
 #include "libc/intrin/atomic.h"
 #include "libc/intrin/strace.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/mem/mem.h"
 #include "libc/nt/console.h"
 #include "libc/nt/enum/filetype.h"

libc/calls/pread.c

@@ -26,7 +26,7 @@
 #include "libc/dce.h"
 #include "libc/intrin/strace.h"
 #include "libc/intrin/weaken.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/runtime/runtime.h"
 #include "libc/runtime/zipos.internal.h"
 #include "libc/stdio/sysparam.h"

libc/calls/program_executable_name_init.S

@@ -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"
 
 	.init.start 305,_init_program_executable_name
 	push	%rdi

libc/calls/pwrite.c

@@ -26,7 +26,7 @@
 #include "libc/calls/syscall-sysv.internal.h"
 #include "libc/dce.h"
 #include "libc/intrin/strace.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/stdio/sysparam.h"
 #include "libc/sysv/errfuns.h"
 

libc/calls/read-nt.c

@@ -16,6 +16,7 @@
 │ TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR             │
 │ PERFORMANCE OF THIS SOFTWARE.                                                │
 ╚─────────────────────────────────────────────────────────────────────────────*/
+#include "libc/assert.h"
 #include "libc/calls/createfileflags.internal.h"
 #include "libc/calls/internal.h"
 #include "libc/calls/sig.internal.h"
@@ -33,7 +34,7 @@
 #include "libc/intrin/nomultics.h"
 #include "libc/intrin/strace.h"
 #include "libc/intrin/weaken.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/nt/console.h"
 #include "libc/nt/createfile.h"
 #include "libc/nt/enum/accessmask.h"
@@ -155,11 +156,11 @@ static textwindows void FreeKeystrokeImpl(struct Dll *key) {
 
 static textwindows struct Keystroke *NewKeystroke(void) {
   struct Dll *e = dll_first(__keystroke.free);
+  if (!e)  // See MIN(freekeys) before ReadConsoleInput()
+    __builtin_trap();
   struct Keystroke *k = KEYSTROKE_CONTAINER(e);
   dll_remove(&__keystroke.free, &k->elem);
   --__keystroke.freekeys;
-  // TODO(jart): What's wrong with GCC 12.3?
-  asm("" : "+r"(k));
   k->buflen = 0;
   return k;
 }
@@ -560,18 +561,15 @@ static textwindows void IngestConsoleInput(void) {
       return;
     if (__keystroke.end_of_file)
       return;
-    if (!GetNumberOfConsoleInputEvents(__keystroke.cin, &n)) {
+    if (!GetNumberOfConsoleInputEvents(__keystroke.cin, &n))
       goto UnexpectedEof;
-    }
     if (!n)
       return;
     n = MIN(__keystroke.freekeys, MIN(ARRAYLEN(records), n));
-    if (!ReadConsoleInput(__keystroke.cin, records, n, &n)) {
+    if (!ReadConsoleInput(__keystroke.cin, records, n, &n))
       goto UnexpectedEof;
-    }
-    for (i = 0; i < n && !__keystroke.end_of_file; ++i) {
+    for (i = 0; i < n && !__keystroke.end_of_file; ++i)
       IngestConsoleInputRecord(records + i);
-    }
   }
 UnexpectedEof:
   STRACE("console read error %d", GetLastError());

libc/calls/readv-metal.c

@@ -22,7 +22,7 @@
 #include "libc/calls/struct/iovec.h"
 #include "libc/calls/struct/iovec.internal.h"
 #include "libc/intrin/weaken.h"
-#include "libc/macros.internal.h"
+#include "libc/macros.h"
 #include "libc/str/str.h"
 #include "libc/sysv/errfuns.h"
 #include "libc/vga/vga.internal.h"

libc/calls/readwrite-nt.c

@@ -19,9 +19,9 @@
 #include "libc/calls/createfileflags.internal.h"
 #include "libc/calls/internal.h"
 #include "libc/calls/sig.internal.h"
-#include "libc/intrin/fds.h"
 #include "libc/calls/struct/sigset.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/errors.h"
@@ -51,39 +51,41 @@ sys_readwrite_nt(int fd, void *data, size_t size, ssize_t offset,
   uint32_t exchanged;
   struct Fd *f = g_fds.p + fd;
 
-  // win32 i/o apis generally take 32-bit values thus we implicitly
-  // truncate outrageously large sizes. linux actually does it too!
-  size = MIN(size, 0x7ffff000);
-
   // pread() and pwrite() perform an implicit lseek() operation, so
   // similar to the lseek() system call, they too raise ESPIPE when
   // operating on a non-seekable file.
   bool pwriting = offset != -1;
-  bool seekable =
-      (f->kind == kFdFile && GetFileType(handle) == kNtFileTypeDisk) ||
-      f->kind == kFdDevNull || f->kind == kFdDevRandom;
-  if (pwriting && !seekable) {
+  bool isdisk = f->kind == kFdFile && GetFileType(handle) == kNtFileTypeDisk;
+  bool seekable = isdisk || f->kind == kFdDevNull || f->kind == kFdDevRandom;
+  if (pwriting && !seekable)
     return espipe();
-  }
 
+  // determine if we need to lock a file descriptor across processes
+  bool locked = isdisk && !pwriting && f->cursor;
+  if (locked)
+    __cursor_lock(f->cursor);
+
+RestartOperation:
   // when a file is opened in overlapped mode win32 requires that we
   // take over full responsibility for managing our own file pointer
   // which is fine, because the one win32 has was never very good in
   // the sense that it behaves so differently from linux, that using
   // win32 i/o required more compatibilty toil than doing it by hand
   if (!pwriting) {
-    if (seekable) {
-      offset = f->pointer;
+    if (seekable && f->cursor) {
+      offset = f->cursor->shared->pointer;
     } else {
       offset = 0;
     }
   }
 
-RestartOperation:
   bool eagained = false;
   // check for signals and cancelation
-  if (_check_cancel() == -1)
+  if (_check_cancel() == -1) {
+    if (locked)
+      __cursor_unlock(f->cursor);
     return -1;  // ECANCELED
+  }
   if (_weaken(__sig_get) && (sig = _weaken(__sig_get)(waitmask))) {
     goto HandleInterrupt;
   }
@@ -114,16 +116,16 @@ RestartOperation:
     }
     ok = true;
   }
-  if (ok) {
+  if (ok)
     ok = GetOverlappedResult(handle, &overlap, &exchanged, true);
-  }
   CloseHandle(overlap.hEvent);
 
   // if i/o succeeded then return its result
   if (ok) {
-    if (!pwriting && seekable) {
-      f->pointer = offset + exchanged;
-    }
+    if (!pwriting && seekable && f->cursor)
+      f->cursor->shared->pointer = offset + exchanged;
+    if (locked)
+      __cursor_unlock(f->cursor);
     return exchanged;
   }
 
@@ -131,23 +133,32 @@ RestartOperation:
   if (GetLastError() == kNtErrorOperationAborted) {
     // raise EAGAIN if it's due to O_NONBLOCK mmode
     if (eagained) {
+      if (locked)
+        __cursor_unlock(f->cursor);
       return eagain();
     }
     // otherwise it must be due to a kill() via __sig_cancel()
     if (_weaken(__sig_relay) && (sig = _weaken(__sig_get)(waitmask))) {
     HandleInterrupt:
+      if (locked)
+        __cursor_unlock(f->cursor);
       int handler_was_called = _weaken(__sig_relay)(sig, SI_KERNEL, waitmask);
       if (_check_cancel() == -1)
         return -1;  // possible if we SIGTHR'd
+      if (locked)
+        __cursor_lock(f->cursor);
       // read() is @restartable unless non-SA_RESTART hands were called
-      if (!(handler_was_called & SIG_HANDLED_NO_RESTART)) {
+      if (!(handler_was_called & SIG_HANDLED_NO_RESTART))
         goto RestartOperation;
-      }
     }
+    if (locked)
+      __cursor_unlock(f->cursor);
     return eintr();
   }
 
   // read() and write() have generally different error-handling paths
+  if (locked)
+    __cursor_unlock(f->cursor);
   return -2;
 }
 

libc/calls/releasefd.c

@@ -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/str/str.h"
 
 // really want to avoid locking here so close() needn't block signals

libc/calls/restore.S

@@ -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"
 .text.windows
 
 //	Restores thread to state before signal.