Remove some legacy cruft

Function trace logs will report stack usage accurately. It won't include
the argv/environ block. Our clone() polyfill is now simpler and does not
use as much stack memory. Function call tracing on x86 is now faster too

libc/dlopen/dlopen.c

@@ -254,7 +254,7 @@ static bool elf_slurp(struct Loaded *l, int fd, const char *file) {
   return true;
 }
 
-static dontinline bool elf_load(struct Loaded *l, const char *file, long pagesz,
+dontinline static bool elf_load(struct Loaded *l, const char *file, long pagesz,
                                 char *interp_path, size_t interp_size) {
   int fd;
   if ((fd = open(file, O_RDONLY | O_CLOEXEC)) == -1)
@@ -280,15 +280,15 @@ static long *push_strs(long *sp, char **list, int count) {
   return sp;
 }
 
-static wontreturn dontinstrument void foreign_helper(void **p) {
+wontreturn dontinstrument static void foreign_helper(void **p) {
   __foreign.dlopen = p[0];
   __foreign.dlsym = p[1];
   __foreign.dlclose = p[2];
   __foreign.dlerror = p[3];
   _longjmp(__foreign.jb, 1);
 }
 
-static dontinline void elf_exec(const char *file, char **envp) {
+dontinline static void elf_exec(const char *file, char **envp) {
 
   // get microprocessor page size
   long pagesz = __pagesize;
@@ -412,7 +412,7 @@ static char *dlerror_set(const char *str) {
   return dlerror_buf;
 }
 
-static dontinline char *foreign_alloc_block(void) {
+dontinline static char *foreign_alloc_block(void) {
   char *p = 0;
   size_t sz = 65536;
   if (!IsWindows()) {
@@ -435,7 +435,7 @@ static dontinline char *foreign_alloc_block(void) {
   return p;
 }
 
-static dontinline void *foreign_alloc(size_t n) {
+dontinline static void *foreign_alloc(size_t n) {
   void *res;
   static char *block;
   __dlopen_lock();
@@ -548,7 +548,7 @@ static void *foreign_thunk_nt(void *func) {
   return code;
 }
 
-static dontinline bool foreign_compile(char exe[hasatleast PATH_MAX]) {
+dontinline static bool foreign_compile(char exe[hasatleast PATH_MAX]) {
 
   // construct path
   strlcpy(exe, get_tmp_dir(), PATH_MAX);

libc/fmt/internal.h

@@ -49,6 +49,6 @@
 int __vcscanf(int (*)(void *), int (*)(int, void *), void *, const char *,
               va_list);
 int __fmt(void *, void *, const char *, va_list, int *);
-__msabi char16_t *__itoa16(char16_t[21], uint64_t);
+char16_t *__itoa16(char16_t[21], uint64_t) __msabi;
 
 #endif /* COSMOPOLITAN_LIBC_FMT_STRTOL_H_ */

libc/intrin/sig.c

@@ -89,14 +89,14 @@ __msabi extern typeof(WriteFile) *const __imp_WriteFile;
 
 extern pthread_mutex_t __sig_worker_lock;
 
-HAIRY static bool __sig_ignored_by_default(int sig) {
+textwindows static bool __sig_ignored_by_default(int sig) {
   return sig == SIGURG ||   //
          sig == SIGCONT ||  //
          sig == SIGCHLD ||  //
          sig == SIGWINCH;
 }
 
-HAIRY bool __sig_ignored(int sig) {
+textwindows bool __sig_ignored(int sig) {
   return __sighandrvas[sig] == (intptr_t)SIG_IGN ||
          (__sighandrvas[sig] == (intptr_t)SIG_DFL &&
           __sig_ignored_by_default(sig));
@@ -532,14 +532,14 @@ textwindows void __sig_generate(int sig, int sic) {
   }
 }
 
-HAIRY static char *__sig_stpcpy(char *d, const char *s) {
+textwindows static char *__sig_stpcpy(char *d, const char *s) {
   size_t i;
   for (i = 0;; ++i)
     if (!(d[i] = s[i]))
       return d + i;
 }
 
-HAIRY wontreturn static void __sig_death(int sig, const char *thing) {
+textwindows wontreturn static void __sig_death(int sig, const char *thing) {
 #ifndef TINY
   intptr_t hStderr;
   char sigbuf[21], s[128], *p;
@@ -810,7 +810,7 @@ HAIRY static uint32_t __sig_worker(void *arg) {
     _pthread_mutex_unlock(&__sig_worker_lock);
     Sleep(POLL_INTERVAL_MS);
   }
-  return 0;
+  __builtin_unreachable();
 }
 
 __attribute__((__constructor__(10))) textstartup void __sig_init(void) {

libc/intrin/sigproc.c

@@ -34,6 +34,8 @@
 #include "libc/nt/thunk/msabi.h"
 #ifdef __x86_64__
 
+#define ABI __msabi textwindows dontinstrument
+
 // cut back on code size and avoid setting errno
 // this code is a mandatory dependency of winmain
 __msabi extern typeof(CloseHandle) *const __imp_CloseHandle;
@@ -47,8 +49,8 @@ __msabi extern typeof(GetEnvironmentVariable)
     *const __imp_GetEnvironmentVariableW;
 
 // Generates C:\ProgramData\cosmo\sig\x\y.pid like path
-__msabi textwindows dontinstrument char16_t *__sig_process_path(
-    char16_t *path, uint32_t pid, int create_directories) {
+ABI char16_t *__sig_process_path(char16_t *path, uint32_t pid,
+                                 int create_directories) {
   char16_t buf[3];
   char16_t *p = path;
   uint32_t vlen = __imp_GetEnvironmentVariableW(u"SYSTEMDRIVE", buf, 3);
@@ -100,7 +102,7 @@ __msabi textwindows dontinstrument char16_t *__sig_process_path(
   return path;
 }
 
-__msabi textwindows atomic_ulong *__sig_map_process(int pid, int disposition) {
+ABI atomic_ulong *__sig_map_process(int pid, int disposition) {
   char16_t path[128];
   __sig_process_path(path, pid, disposition == kNtOpenAlways);
   intptr_t hand = __imp_CreateFileW(path, kNtGenericRead | kNtGenericWrite,

libc/intrin/sigprocmask-sysv.c

@@ -32,8 +32,7 @@ int sys_sigprocmask(int how, const sigset_t *opt_set,
         how, opt_set ? (sigset_t *)(intptr_t)(uint32_t)*opt_set : 0, 0, 0);
     rc = 0;
   }
-  if (rc != -1 && opt_out_oldset) {
+  if (rc != -1 && opt_out_oldset)
     *opt_out_oldset = old[0];
-  }
   return rc;
 }

libc/intrin/strsignal_r.c

@@ -36,12 +36,10 @@
 privileged const char *strsignal_r(int sig, char buf[21]) {
   char *p;
   const char *s;
-  if (!sig) {
+  if (!sig)
     return "0";
-  }
-  if ((s = GetMagnumStr(kSignalNames, sig))) {
+  if ((s = GetMagnumStr(kSignalNames, sig)))
     return s;
-  }
   if (SIGRTMIN <= sig && sig <= SIGRTMAX) {
     sig -= SIGRTMIN;
     buf[0] = 'S';

libc/intrin/ulock.c

@@ -79,7 +79,11 @@ int ulock_wait(uint32_t operation, void *addr, uint64_t value,
 //   it could also mean another thread calling ulock on this address was
 //   configured (via operation) in an inconsistent way.
 //
-int ulock_wake(uint32_t operation, void *addr, uint64_t wake_value) {
+// should be dontinstrument because SiliconThreadMain() calls this from
+// a stack managed by apple libc.
+//
+dontinstrument int ulock_wake(uint32_t operation, void *addr,
+                              uint64_t wake_value) {
   int rc;
   rc = __syscall3i(operation, (long)addr, wake_value, 0x2000000 | 516);
   LOCKTRACE("ulock_wake(%#x, %p, %lx) → %s", operation, addr, wake_value,

libc/log/backtrace3.c

@@ -48,9 +48,8 @@
  * @param st is open symbol table for current executable
  * @return -1 w/ errno if error happened
  */
-dontinstrument int PrintBacktraceUsingSymbols(int fd,
-                                              const struct StackFrame *bp,
-                                              struct SymbolTable *st) {
+int PrintBacktraceUsingSymbols(int fd, const struct StackFrame *bp,
+                               struct SymbolTable *st) {
   size_t gi;
   char *cxxbuf;
   intptr_t addr;

libc/log/watch.c

@@ -33,14 +33,14 @@ static char __watch_last[4096];
 
 void __watch_hook(void);
 
-static dontinstrument inline void Copy(char *p, char *q, size_t n) {
+dontinstrument static inline void Copy(char *p, char *q, size_t n) {
   size_t i;
   for (i = 0; i < n; ++i) {
     p[i] = q[i];
   }
 }
 
-static dontinstrument inline int Cmp(char *p, char *q, size_t n) {
+dontinstrument static inline int Cmp(char *p, char *q, size_t n) {
   if (n == 8)
     return READ64LE(p) != READ64LE(q);
   if (n == 4)

libc/macros.h

@@ -291,21 +291,6 @@
 	.balign	4
 .endm
 
-//	Loads address of errno into %rcx
-.macro	.errno
-	call	__errno_location
-.endm
-
-//	Post-Initialization Read-Only (PIRO) BSS section.
-//	@param	ss is an optional string, for control image locality
-.macro	.piro	ss
- .ifnb	\ss
-	.section .piro.sort.bss.\ss,"aw",@nobits
- .else
-	.section .piro.bss,"aw",@nobits
- .endif
-.endm
-
 //	Helpers for Cosmopolitan _init() amalgamation magic.
 //	@param	name should be consistent across macros for a module
 //	@see	libc/runtime/_init.S

libc/proc/proc.c

@@ -71,7 +71,7 @@ struct Procs __proc = {
     .lock = PTHREAD_MUTEX_INITIALIZER,
 };
 
-static textwindows void __proc_stats(int64_t h, struct rusage *ru) {
+textwindows static void __proc_stats(int64_t h, struct rusage *ru) {
   bzero(ru, sizeof(*ru));
   struct NtProcessMemoryCountersEx memcount = {sizeof(memcount)};
   GetProcessMemoryInfo(h, &memcount, sizeof(memcount));
@@ -137,7 +137,7 @@ textwindows int __proc_harvest(struct Proc *pr, bool iswait4) {
   return sic;
 }
 
-static textwindows dontinstrument uint32_t __proc_worker(void *arg) {
+textwindows dontinstrument static uint32_t __proc_worker(void *arg) {
   struct CosmoTib tls;
   char *sp = __builtin_frame_address(0);
   __bootstrap_tls(&tls, __builtin_frame_address(0));
@@ -246,7 +246,7 @@ static textwindows dontinstrument uint32_t __proc_worker(void *arg) {
 /**
  * Lazy initializes process tracker data structures and worker.
  */
-static textwindows void __proc_setup(void) {
+textwindows static void __proc_setup(void) {
   __proc.onbirth = CreateEvent(0, 0, 0, 0);     // auto reset
   __proc.haszombies = CreateEvent(0, 1, 0, 0);  // manual reset
   __proc.thread = CreateThread(0, STACK_SIZE, __proc_worker, 0,

libc/runtime/clone-linux.S

@@ -26,7 +26,7 @@
 //	@param	rdx	x2	is ptid
 //	@param	rcx	x3	is ctid
 //	@param	r8	x4	is tls
-//	@param	r9	x5	is func(void*,int)→int
+//	@param	r9	x5	is func(void*)→int
 //	@param	8(rsp)	x6	is arg
 //	@return	tid of child on success, or -errno on error
 sys_clone_linux:
@@ -45,16 +45,10 @@ sys_clone_linux:
 	ret
 2:	xor	%ebp,%ebp		// child thread
 	mov	%rbx,%rdi		// arg
-	mov	%r10,%r15		// experiment
-	mov	(%r10),%esi		// tid
 	call	*%r9			// func(arg,tid)
 	xchg	%eax,%edi		// func(arg,tid) → exitcode
-	mov	(%r15),%eax		// experiment
-	test	%eax,%eax		// experiment
-	jz	1f			// experiment
 	mov	$60,%eax		// __NR_exit(exitcode)
 	syscall
-1:	hlt				// ctid was corrupted by program!
 #elif defined(__aarch64__)
 	stp	x29,x30,[sp,#-16]!
 	mov	x29,sp
@@ -69,7 +63,6 @@ sys_clone_linux:
 2:	mov	x29,#0			// wipe backtrace
 	mov	x28,x3			// set cosmo tls
 	mov	x0,x6			// child thread
-	ldr	w1,[x4]			// arg2 = *ctid
 	blr	x5
 	mov	x8,#93			// __NR_exit
 	svc	#0