tccgen: void values, etc...

originally, this was meant to avoid 'void' values leak
to the gen-xxx.c generators which would otherwise try
to load void from/to registers.  Also catch invalid types
such as 'unsigned bool' eariler.

tccgen.c:
- expr_cond()/gexpr(): ignore VT_VOID
- gen_cast(): set float reg with static initializers to make
  backends happier with invalid input
- init_putv():
  improve static compound literal initializers
Also:
- ignore "missing prototype" and "might return no value"
  for old style K&R functions

tcc.c, libtcc.c:
- tcc_set_output_type(): return errors from loading crt1.o

tcc.h/tccpp.c:
- fix horrible longstanding mistake with sizeof SValue.tab.
  Must have place for sizeof (long double) rather than the
  target's LDOUBLE_SIZE.

tccdefs.h:
- add a fake __[u]int128_t for all platforms

arm-gen.c:
- gen_cvt_ftof() must load to reg always

arm64-gen.c:
- simplify arm64_pcs() a bit

i386/x86_64-gen.c:
- do not assume char is signed (for cross compilation arm->x86)

tccelf/macho linker message:
- instead of
      "undefined symbol 'X'"
  now say
      "unresolved reference to 'X'"
  which makes it more clear that it is a _linker_error_.

tccasm.c:
- .align/p2align: use 'nop's in exec sections

Author: grischka

arm-gen.c

@@ -2283,8 +2283,9 @@ void gen_cvt_ftoi(int t)
 void gen_cvt_ftof(int t)
 {
 #ifdef TCC_ARM_VFP
+  uint32_t r = gv(RC_FLOAT);
   if(((vtop->type.t & VT_BTYPE) == VT_FLOAT) != ((t & VT_BTYPE) == VT_FLOAT)) {
-    uint32_t r = vfpr(gv(RC_FLOAT));
+    r = vfpr(r);
     o(0xEEB70AC0|(r<<12)|r|T2CPR(vtop->type.t));
   }
 #else

arm64-gen.c

@@ -278,7 +278,6 @@ static int arm64_type_size(int t)
     case VT_DOUBLE: return 3;
     case VT_LDOUBLE: return 4;
     case VT_BOOL: return 0;
-    case VT_VOID: return 0;
     }
     assert(0);
     return 0;
@@ -832,11 +831,10 @@ static unsigned long arm64_pcs_aux(int variadic, int n, CType **type, unsigned l
 
     for (i = 0; i < n; i++) {
         int hfa = arm64_hfa(type[i], 0);
-        int win_vararg_float = 0;
-        int size, align;
+        int size, align, bt;
 
-        if ((type[i]->t & VT_ARRAY) ||
-            (type[i]->t & VT_BTYPE) == VT_FUNC)
+        bt = type[i]->t & VT_BTYPE;
+        if (bt == VT_PTR || bt == VT_FUNC)
             size = align = 8;
         else
             size = type_size(type[i], &align);
@@ -848,13 +846,10 @@ static unsigned long arm64_pcs_aux(int variadic, int n, CType **type, unsigned l
 	}
 
 #elif defined(TCC_TARGET_PE)
-        if (variadic && i >= variadic && (hfa || is_float(type[i]->t))) {
+        if (variadic && i >= variadic) {
             hfa = 0;
-            if (is_float(type[i]->t)) {
-                win_vararg_float = 1;
-                size = 8;
-                align = 8;
-            }
+            if (is_float(bt))
+                bt = VT_INT, size = align = 8;
         }
 #endif
         if (hfa)
@@ -871,12 +866,12 @@ static unsigned long arm64_pcs_aux(int variadic, int n, CType **type, unsigned l
             }
             continue;
         }
-        else if ((type[i]->t & VT_BTYPE) == VT_STRUCT)
+        else if (bt == VT_STRUCT)
             // B.4
             size = (size + 7) & ~7;
 
         // C.1
-        if (!win_vararg_float && is_float(type[i]->t) && nv < 8) {
+        if (is_float(bt) && nv < 8) {
             a[i] = 16 + (nv++ << 1);
             continue;
         }
@@ -895,24 +890,24 @@ static unsigned long arm64_pcs_aux(int variadic, int n, CType **type, unsigned l
         }
 
         // C.4
-        if (hfa || (type[i]->t & VT_BTYPE) == VT_LDOUBLE) {
+        if (hfa || bt == VT_LDOUBLE) {
             ns = (ns + 7) & ~7;
             ns = (ns + align - 1) & -align;
         }
 
         // C.5
-        if ((type[i]->t & VT_BTYPE) == VT_FLOAT)
+        if (bt == VT_FLOAT)
             size = 8;
 
         // C.6
-        if (!win_vararg_float && (hfa || is_float(type[i]->t))) {
+        if (hfa || is_float(bt)) {
             a[i] = ns;
             ns += size;
             continue;
         }
 
         // C.7
-        if ((type[i]->t & VT_BTYPE) != VT_STRUCT && size <= 8 && nx < 8) {
+        if (bt != VT_STRUCT && size <= 8 && nx < 8) {
             a[i] = nx++ << 1;
             continue;
         }
@@ -922,14 +917,14 @@ static unsigned long arm64_pcs_aux(int variadic, int n, CType **type, unsigned l
             nx = (nx + 1) & ~1;
 
         // C.9
-        if ((type[i]->t & VT_BTYPE) != VT_STRUCT && size == 16 && nx < 7) {
+        if (bt != VT_STRUCT && size == 16 && nx < 7) {
             a[i] = nx << 1;
             nx += 2;
             continue;
         }
 
         // C.10
-        if ((type[i]->t & VT_BTYPE) == VT_STRUCT && size <= (8 - nx) * 8) {
+        if (bt == VT_STRUCT && size <= (8 - nx) * 8) {
             a[i] = nx << 1;
             nx += (size + 7) >> 3;
             continue;
@@ -943,7 +938,7 @@ static unsigned long arm64_pcs_aux(int variadic, int n, CType **type, unsigned l
         ns = (ns + align - 1) & -align;
 
         // C.13
-        if ((type[i]->t & VT_BTYPE) == VT_STRUCT) {
+        if (bt == VT_STRUCT) {
             a[i] = ns;
             ns += size;
             continue;
@@ -974,11 +969,11 @@ static unsigned long arm64_pcs(int variadic, int n, CType **type, unsigned long
     }
 
     // Argument types:
-    stack = arm64_pcs_aux(variadic, n, type + 1, a + 1);
+    stack = arm64_pcs_aux(variadic, n - 1, type + 1, a + 1);
 
     if (0) {
         int i;
-        for (i = 0; i <= n; i++) {
+        for (i = 0; i < n; i++) {
             if (!i)
                 printf("arm64_pcs return: ");
             else
@@ -1076,9 +1071,9 @@ ST_FUNC void gfunc_call(int nb_args)
 
     stack = arm64_pcs(
 #ifdef TCC_TARGET_PE
-        old_style ? -1 :
+        old_style ? /* all args like varargs in K&R style */ -1 :
 #endif
-        var_nb_arg, nb_args, t, a);
+        var_nb_arg, nb_args + 1, t, a);
 
     // Allocate space for structs replaced by pointer:
     for (i = nb_args; i; i--)
@@ -1258,43 +1253,39 @@ ST_FUNC void gfunc_prolog(Sym *func_sym)
     CType *func_type = &func_sym->type;
     int n = 0;
     int i = 0;
-    int pcs_n;
     Sym *sym;
     CType **t;
     unsigned long *a;
     int use_x8 = 0;
     int last_int = 0;
     int last_float = 0;
     int variadic = func_sym->type.ref->f.func_type == FUNC_ELLIPSIS;
-    int var_nb_arg = n_func_args(&func_sym->type);
-    int c;
+    int var_nb_arg = variadic ? n_func_args(&func_sym->type) : 0;
 
     func_vc = 144; // offset of where x8 is stored
 
     for (sym = func_type->ref; sym; sym = sym->next)
         ++n;
 
-    pcs_n = n - 1;
-    c = n + variadic;
-    t = tcc_malloc(c * sizeof(*t));
-    a = tcc_malloc(c * sizeof(*a));
+#ifdef TCC_TARGET_PE
+    n += variadic;
+#endif
 
+    t = tcc_malloc(n * sizeof(*t));
+    a = tcc_malloc(n * sizeof(*a));
     for (sym = func_type->ref; sym; sym = sym->next)
         t[i++] = &sym->type;
 
 #ifdef TCC_TARGET_PE
-    if (variadic) {
-        t[i++] = &int_type;
-        ++pcs_n;
-    }
+    if (variadic)
+        t[i] = &int_type;
 #endif
 
-    arm64_func_va_list_stack = arm64_pcs(variadic ? var_nb_arg : 0,
-                                         pcs_n, t, a);
+    arm64_func_va_list_stack = arm64_pcs(var_nb_arg, n, t, a);
 
 #ifdef TCC_TARGET_PE
     if (variadic)
-        arm64_func_va_list_stack = arm64_pe_param_off(a[n]);
+        arm64_func_va_list_stack = arm64_pe_param_off(a[n - 1]);
 #endif
 
 #if !defined(TCC_TARGET_MACHO)
@@ -1304,6 +1295,7 @@ ST_FUNC void gfunc_prolog(Sym *func_sym)
         last_float = 4;
     }
 #endif
+
     if (a && a[0] == 1)
         use_x8 = 1;
     for (i = 1, sym = func_type->ref->next; sym; i++, sym = sym->next) {
@@ -1563,7 +1555,8 @@ ST_FUNC void gfunc_return(CType *func_type)
     CType *t = func_type;
     unsigned long a;
 
-    arm64_pcs(0, 0, &t, &a);
+    arm64_pcs(0, 1, &t, &a);
+
     switch (a) {
     case -1:
         break;

i386-gen.c

@@ -271,7 +271,7 @@ static void gen_modrm(int opc, int op_r2, int r, Sym *sym, int c)
     } else if ((r & VT_VALMASK) == VT_LOCAL) {
 	o(opc);
         /* currently, we use only ebp as base */
-        if (c == (char)c) {
+        if (c == (signed char)c) {
             /* short reference */
             o(0x45 | op_reg);
             g(c);
@@ -439,7 +439,7 @@ ST_FUNC void store(int r, SValue *v)
 
 static void gadd_sp(int val)
 {
-    if (val == (char)val) {
+    if (val == (signed char)val) {
         o(0xc483);
         g(val);
     } else {
@@ -774,7 +774,7 @@ ST_FUNC void gjmp_addr(int a)
 {
     int r;
     r = a - ind - 2;
-    if (r == (char)r) {
+    if (r == (signed char)r) {
         g(0xeb);
         g(r);
     } else {
@@ -787,7 +787,7 @@ ST_FUNC void gjmp_addr(int a)
 ST_FUNC void gjmp_cond_addr(int a, int op)
 {
     int r = a - ind - 2;
-    if (r == (char)r)
+    if (r == (signed char)r)
         g(op - 32), g(r);
     else
         g(0x0f), gjmp2(op - 16, r - 4);
@@ -830,7 +830,7 @@ ST_FUNC void gen_opi(int op)
             r = gv(RC_INT);
             vswap();
             c = vtop->c.i;
-            if (c == (char)c) {
+            if (c == (signed char)c) {
                 /* generate inc and dec for smaller code */
                 if ((c == 1 || c == -1) && (op == '+' || op == '-')) {
                     opc = (c == 1) ^ (op == '+');

include/tccdefs.h

@@ -90,8 +90,6 @@
     #define __NO_TLS 1
     #define __RUNETYPE_INTERNAL 1
 # if __SIZEOF_POINTER__ == 8
-    /* FIXME, __int128_t is used by setjump */
-    #define __int128_t struct { unsigned char _dummy[16] __attribute((aligned(16))); }
     #define __SIZEOF_SIZE_T__ 8
     #define __SIZEOF_PTRDIFF_T__ 8
 #else
@@ -142,12 +140,6 @@
 #endif
     #define __INT32_TYPE__ int
 
-#if defined __aarch64__
-    /* GCC's __uint128_t appears in some Linux/OSX header files. Make it a
-       synonym for long double to get the size and alignment right. */
-    #define __uint128_t long double
-#endif
-
 #if !defined _WIN32
     /* glibc defines. We do not support __USER_NAME_PREFIX__ */
     #define __REDIRECT(name, proto, alias) name proto __asm__ (#alias)
@@ -187,6 +179,12 @@
 # endif
 #endif
 
+    /* GCC's __uint128_t appears in some Linux/OSX header files.
+       Just make it some type with same size and alignment. */
+    struct __uint128__ { char x[16]; } __attribute((__aligned__(16)));
+    #define __int128_t struct __uint128__
+    #define __uint128_t struct __uint128__
+
     /* __builtin_va_list */
 #if defined __x86_64__
 #if !defined _WIN32

libtcc.c

@@ -252,7 +252,7 @@ static void *default_reallocator(void *ptr, unsigned long size)
     else {
         ptr1 = realloc(ptr, size);
         if (!ptr1) {
-            fprintf(stderr, "memory full\n");
+            fprintf(stderr, "tcc: memory full\n");
             exit (1);
         }
     }
@@ -1003,9 +1003,9 @@ LIBTCCAPI int tcc_set_output_type(TCCState *s, int output_type)
     /* paths for crt objects */
     tcc_split_path(s, &s->crt_paths, &s->nb_crt_paths, CONFIG_TCC_CRTPREFIX);
     if (output_type != TCC_OUTPUT_MEMORY && !s->nostdlib)
-        tccelf_add_crtbegin(s);
+        tccelf_add_crtbegin(s); /* may produce errors */
 #endif
-    return 0;
+    return s->nb_errors ? -1 : 0;
 }
 
 LIBTCCAPI int tcc_add_include_path(TCCState *s, const char *pathname)

tcc.c

@@ -35,7 +35,7 @@ static const char help[] =
     "General options:\n"
     "  -c           compile only - generate an object file\n"
     "  -o outfile   set output filename\n"
-    "  -run         run compiled source [with custom stdin: -rstdin FILE]\n"
+    "  -run         run compiled source\n"
     "  -fflag       set or reset (with 'no-' prefix) 'flag' (see tcc -hh)\n"
     "  -Wwarning    set or reset (with 'no-' prefix) 'warning' (see tcc -hh)\n"
     "  -w           disable all warnings\n"
@@ -105,6 +105,7 @@ static const char help2[] =
     "  -static                       link to static libraries (not recommended)\n"
     "  -dumpversion                  print version\n"
     "  -print-search-dirs            print search paths\n"
+    "  -rstdin file                  with -run: use 'file' as custom stdin\n"
     "  -dt                           with -run/-E: auto-define 'test_...' macros\n"
     "Ignored options:\n"
     "  -arch -C --param -pedantic -pipe -s -traditional\n"
@@ -354,9 +355,10 @@ int main(int argc, char **argv)
     set_environment(s);
     if (s->output_type == 0)
         s->output_type = TCC_OUTPUT_EXE;
-    tcc_set_output_type(s, s->output_type);
+    ret = tcc_set_output_type(s, s->output_type);
     if (ppfp)
         s->ppfp = ppfp;
+
     if ((s->output_type == TCC_OUTPUT_MEMORY
       || s->output_type == TCC_OUTPUT_PREPROCESS)
         && (s->dflag & 16)) { /* -dt option */
@@ -369,7 +371,7 @@ int main(int argc, char **argv)
 
     /* compile or add each files or library */
     first_file = NULL;
-    do {
+    while (0 == ret) {
         struct filespec *f = s->files[n];
         s->filetype = f->type;
         if (f->type & AFF_TYPE_LIB) {
@@ -381,9 +383,11 @@ int main(int argc, char **argv)
                 first_file = f->name;
             ret = tcc_add_file(s, f->name);
         }
-    } while (++n < s->nb_files
-            && 0 == ret
-            && (s->output_type != TCC_OUTPUT_OBJ || s->option_r));
+        if (++n == s->nb_files)
+            break;
+        if (s->output_type == TCC_OUTPUT_OBJ && !s->option_r)
+            break;
+    }
 
     if (s->do_bench)
         end_time = getclock_ms();