win32: pe dwarf sections & option -g.pdb

On windows, create a .pdb file with option "-g.pdb"
Such executables created by tcc can be debugged with
"ollydbg" or "x64dbg"

This currently relies on the 3rd party tool cv2pdb from
    https://github.com/rainers/cv2pdb
which again relies on
    mspdbsrv.exe mspdbcore.dll msobj80.dll mspdb80.dll
from a MSVC installation.

cv2pdb.exe + the ms* files may be put in the path or in the
same directory as tcc.exe.

Author: grischka

libtcc.c

@@ -1976,9 +1976,12 @@ PUB_FUNC int tcc_parse_args(TCCState *s, int *pargc, char ***pargv, int optind)
         case TCC_OPTION_g:
             s->do_debug = 1;
             s->dwarf = DWARF_VERSION;
-
             if (strstart("dwarf", &optarg))
                 s->dwarf = (*optarg) ? (0 - atoi(optarg)) : DEFAULT_DWARF_VERSION;
+#ifdef TCC_TARGET_PE
+            else if (0 == strcmp(".pdb", optarg))
+                s->dwarf = 5, s->do_debug |= 16;
+#endif
             break;
         case TCC_OPTION_c:
             x = TCC_OUTPUT_OBJ;

tcc.c

@@ -59,6 +59,9 @@ static const char help[] =
     "Debugger options:\n"
     "  -g           generate stab runtime debug info\n"
     "  -gdwarf[-x]  generate dwarf runtime debug info\n"
+#ifdef TCC_TARGET_PE
+    "  -g.pdb       create .pdb debug database\n"
+#endif
 #ifdef CONFIG_TCC_BCHECK
     "  -b           compile with built-in memory and bounds checker (implies -g)\n"
 #endif

tccdbg.c

@@ -1086,7 +1086,7 @@ ST_FUNC void tcc_debug_line(TCCState *s1)
 
     if (!s1->do_debug)
         return;
-    if (cur_text_section != text_section)
+    if (cur_text_section != text_section || nocode_wanted)
         return;
     f = put_new_file(s1);
     if (!f)

tccgen.c

@@ -5500,25 +5500,12 @@ ST_FUNC void unary(void)
             goto tok_identifier;
         /* fall thru */
     case TOK___FUNC__:
-        {
-            Section *sec;
-            int len;
-            /* special function name identifier */
-            len = strlen(funcname) + 1;
-            /* generate char[len] type */
-            type.t = char_type.t;
-            if (tcc_state->warn_write_strings & WARN_ON)
-                type.t |= VT_CONSTANT;
-            mk_pointer(&type);
-            type.t |= VT_ARRAY;
-            type.ref->c = len;
-            sec = rodata_section;
-            vpush_ref(&type, sec, sec->data_offset, len);
-            if (!NODATA_WANTED)
-                memcpy(section_ptr_add(sec, len), funcname, len);
-            next();
-        }
-        break;
+        tok = TOK_STR;
+        cstr_reset(&tokcstr);
+        cstr_cat(&tokcstr, funcname, 0);
+        tokc.str.size = tokcstr.size;
+        tokc.str.data = tokcstr.data;
+        goto case_TOK_STR;
     case TOK_LSTR:
 #ifdef TCC_TARGET_PE
         t = VT_SHORT | VT_UNSIGNED;
@@ -5527,6 +5514,7 @@ ST_FUNC void unary(void)
 #endif
         goto str_init;
     case TOK_STR:
+    case_TOK_STR:
         /* string parsing */
         t = char_type.t;
     str_init:
@@ -7767,7 +7755,7 @@ static void decl_initializer(init_params *p, CType *type, unsigned long c, int f
     CType *t1;
 
     /* generate line number info */
-    if (debug_modes && !p->sec)
+    if (debug_modes && !(flags & DIF_SIZE_ONLY) && !p->sec)
         tcc_debug_line(tcc_state), tcc_tcov_check_line (tcc_state, 1);
 
     if (!(flags & DIF_HAVE_ELEM) && tok != '{' &&
@@ -8288,50 +8276,60 @@ static void gen_function(Sym *sym)
     struct scope f = { 0 };
     cur_scope = root_scope = &f;
     nocode_wanted = 0;
+
     ind = cur_text_section->data_offset;
     if (sym->a.aligned) {
 	size_t newoff = section_add(cur_text_section, 0,
 				    1 << (sym->a.aligned - 1));
 	gen_fill_nops(newoff - ind);
     }
+
+    funcname = get_tok_str(sym->v, NULL);
+    func_ind = ind;
+    func_vt = sym->type.ref->type;
+    func_var = sym->type.ref->f.func_type == FUNC_ELLIPSIS;
+
     /* NOTE: we patch the symbol size later */
     put_extern_sym(sym, cur_text_section, ind, 0);
+
     if (sym->type.ref->f.func_ctor)
         add_array (tcc_state, ".init_array", sym->c);
     if (sym->type.ref->f.func_dtor)
         add_array (tcc_state, ".fini_array", sym->c);
 
-    funcname = get_tok_str(sym->v, NULL);
-    func_ind = ind;
-    func_vt = sym->type.ref->type;
-    func_var = sym->type.ref->f.func_type == FUNC_ELLIPSIS;
-
     /* put debug symbol */
     tcc_debug_funcstart(tcc_state, sym);
+
     /* push a dummy symbol to enable local sym storage */
     sym_push2(&local_stack, SYM_FIELD, 0, 0);
     local_scope = 1; /* for function parameters */
     gfunc_prolog(sym);
     tcc_debug_prolog_epilog(tcc_state, 0);
+
     local_scope = 0;
     rsym = 0;
     clear_temp_local_var_list();
     func_vla_arg(sym);
     block(0);
     gsym(rsym);
+
     nocode_wanted = 0;
     /* reset local stack */
     pop_local_syms(NULL, 0);
     tcc_debug_prolog_epilog(tcc_state, 1);
     gfunc_epilog();
+
+    /* end of function */
+    tcc_debug_funcend(tcc_state, ind - func_ind);
+
+    /* patch symbol size */
+    elfsym(sym)->st_size = ind - func_ind;
+
     cur_text_section->data_offset = ind;
     local_scope = 0;
     label_pop(&global_label_stack, NULL, 0);
     sym_pop(&all_cleanups, NULL, 0);
-    /* patch symbol size */
-    elfsym(sym)->st_size = ind - func_ind;
-    /* end of function */
-    tcc_debug_funcend(tcc_state, ind - func_ind);
+
     /* It's better to crash than to generate wrong code */
     cur_text_section = NULL;
     funcname = ""; /* for safety */
@@ -8341,6 +8339,7 @@ static void gen_function(Sym *sym)
     func_ind = -1;
     nocode_wanted = DATA_ONLY_WANTED;
     check_vstack();
+
     /* do this after funcend debug info */
     next();
 }

tccpe.c

@@ -301,8 +301,7 @@ enum {
     sec_pdata ,
     sec_other ,
     sec_rsrc ,
-    sec_stab ,
-    sec_stabstr ,
+    sec_debug ,
     sec_reloc ,
     sec_last
 };
@@ -475,9 +474,9 @@ struct pe_file {
     unsigned pos;
 };
 
-static int pe_fwrite(void *data, int len, struct pe_file *pf)
+static int pe_fwrite(const void *data, int len, struct pe_file *pf)
 {
-    WORD *p = data;
+    const WORD *p = data;
     DWORD sum;
     int ret, i;
     pf->pos += (ret = fwrite(data, 1, len, pf->op));
@@ -502,6 +501,51 @@ static void pe_fpad(struct pe_file *pf, DWORD new_pos)
     pf->pos = new_pos;
 }
 
+/*----------------------------------------------------------------------------*/
+/* PE-DWARF/COFF support
+   does not work with a mingw-gdb really but works with cv2pdb
+   (https://github.com/rainers/cv2pdb) */
+
+#define N_COFF_SYMS 0
+
+static const char dwarf_secs[] =
+{
+    ".debug_info\0"
+    ".debug_abbrev\0"
+    ".debug_line\0"
+    ".debug_aranges\0"
+    ".debug_str\0"
+    ".debug_line_str\0"
+};
+
+static const unsigned coff_strtab_size = 4 + sizeof dwarf_secs - 1;
+
+static int pe_put_long_secname(char *secname, const char *name)
+{
+    const char *d = dwarf_secs;
+    do {
+        if (0 == strcmp(d, name)) {
+            sprintf(secname, "/%d", (int)(d - dwarf_secs + 4));
+            return 1;
+        }
+        d = strchr(d, 0) + 1;
+    } while (*d);
+    return 0;
+}
+
+static void pe_create_pdb(TCCState *s1, const char *exename)
+{
+    char buf[300]; int r;
+    snprintf(buf, sizeof buf, "cv2pdb.exe %s", exename);
+    r = system(buf);
+    strcpy(tcc_fileextension(strcpy(buf, exename)), ".pdb");
+    if (r) {
+        tcc_error_noabort("could not create '%s'\n(need working cv2pdb from https://github.com/rainers/cv2pdb)", buf);
+    } else if (s1->verbose) {
+        printf("<- %s\n", buf);
+    }
+}
+
 /*----------------------------------------------------------------------------*/
 static int pe_write(struct pe_info *pe)
 {
@@ -616,6 +660,7 @@ static int pe_write(struct pe_info *pe)
     struct section_info *si;
     IMAGE_SECTION_HEADER *psh;
     TCCState *s1 = pe->s1;
+    int need_strtab = 0;
 
     pf.op = fopen(pe->filename, "wb");
     if (NULL == pf.op)
@@ -686,6 +731,8 @@ static int pe_write(struct pe_info *pe)
         }
 
         memcpy(psh->Name, sh_name, umin(strlen(sh_name), sizeof psh->Name));
+        if (si->cls == sec_debug)
+            need_strtab += pe_put_long_secname(psh->Name, sh_name);
 
         psh->Characteristics = si->pe_flags;
         psh->VirtualAddress = addr;
@@ -715,7 +762,10 @@ static int pe_write(struct pe_info *pe)
     if (PE_DLL == pe->type)
         pe_header.filehdr.Characteristics = CHARACTERISTICS_DLL;
     pe_header.filehdr.Characteristics |= pe->s1->pe_characteristics;
-
+    if (need_strtab) {
+        pe_header.filehdr.PointerToSymbolTable = file_offset;
+        pe_header.filehdr.NumberOfSymbols = N_COFF_SYMS;
+    }
     pe_fwrite(&pe_header, sizeof pe_header, &pf);
     for (i = 0; i < pe->sec_count; ++i)
         pe_fwrite(&pe->sec_info[i]->ish, sizeof(IMAGE_SECTION_HEADER), &pf);
@@ -736,6 +786,13 @@ static int pe_write(struct pe_info *pe)
         pe_fpad(&pf, file_offset);
     }
 
+    if (need_strtab) {
+        /* create a tiny COFF string table with the long section names */
+        pe_fwrite(&coff_strtab_size, sizeof coff_strtab_size, &pf);
+        pe_fwrite(dwarf_secs, sizeof dwarf_secs - 1, &pf);
+        file_offset = pf.pos;
+    }
+
     pf.sum += file_offset;
     fseek(pf.op, offsetof(struct pe_header, opthdr.CheckSum), SEEK_SET);
     pe_fwrite(&pf.sum, sizeof (DWORD), &pf);
@@ -750,6 +807,8 @@ static int pe_write(struct pe_info *pe)
     if (pe->s1->verbose)
         printf("<- %s (%u bytes)\n", pe->filename, (unsigned)file_offset);
 
+    if (s1->do_debug & 16)
+        pe_create_pdb(s1, pe->filename);
     return 0;
 }
 
@@ -1072,12 +1131,12 @@ static int pe_section_class(Section *s)
     type = s->sh_type;
     flags = s->sh_flags;
     name = s->name;
-    if (0 == memcmp(name, ".stab", 5)) {
-        if (0 == s->s1->do_debug)
-            return sec_last;
-        return name[5] ? sec_stabstr : sec_stab;
-    }
-    if (flags & SHF_ALLOC) {
+
+    if (0 == memcmp(name, ".stab", 5) || 0 == memcmp(name, ".debug_", 7)) {
+        if (s->s1->do_debug)
+            return sec_debug;
+
+    } else if (flags & SHF_ALLOC) {
         if (type == SHT_PROGBITS
          || type == SHT_INIT_ARRAY
          || type == SHT_FINI_ARRAY) {
@@ -1135,7 +1194,7 @@ static int pe_assign_addresses (struct pe_info *pe)
         if (PE_MERGE_DATA && c == sec_bss)
             c = sec_data;
 
-        if (si && c == si->cls) {
+        if (si && c == si->cls && c != sec_debug) {
             /* merge with previous section */
             s->sh_addr = addr = ((addr - 1) | (16 - 1)) + 1;
         } else {