feat: support extern

.gitignore

@@ -17,6 +17,9 @@ tmp*
 
 # for debug
 test
+ext
+test.*
+ext.*
 .gdb_history
 
 # stages

.vscode/settings.json

@@ -7,7 +7,8 @@
     "*.plt": "prolog",
     "stdbool.h": "c",
     "tests": "cpp",
-    "initializer_list": "c"
+    "initializer_list": "c",
+    "string.h": "c"
   },
   "C_Cpp.default.configurationProvider": "ms-vscode.makefile-tools"
 }

Makefile

@@ -2,25 +2,8 @@ CFLAGS=-std=c11 -g -static
 SRCS=$(wildcard *.c)
 OBJS=$(SRCS:.c=.o)
 
-ccc: $(OBJS)
-	$(CC) -o $@ $(OBJS) $(CFLAGS)
-
 $(OBJS): ccc.h
 
-STAGES = stage1 stage2 stage3
-
-define PROCESS_STAGE_RULE
-$(1)/%.c: %.c
-	mkdir -p $(1)
-	gcc -E $$< | grep -v '^#' > $$@
-endef
-
-$(foreach stage,$(STAGES),$(eval $(call PROCESS_STAGE_RULE,$(stage))))
-
-PROCESSED_SRCS = $(foreach stage,$(STAGES),$(addprefix $(stage)/,$(SRCS)))
-
-process: $(PROCESSED_SRCS)
-
 help:
 	@echo "Usage: make [lint|test|debug|clean|process]"
 	@echo "  lint: run c and shell checker"
@@ -29,6 +12,9 @@ help:
 	@echo "  clean: remove object files"
 	@echo "  process: run the preprocessor on all source files for each stage"
 
+ccc: $(OBJS)
+	$(CC) -o $@ $(OBJS) $(CFLAGS)
+
 lint:
 	clang-format -i ./*.c
 	clang-format -i ./*.h
@@ -39,13 +25,42 @@ lintx:
 test: ccc
 	./ccc tests > tmp.s
 	echo 'int char_fn() { return 257; }' | gcc -xc -c -o tmp2.o -
-	gcc -static -o tmp tmp.s tmp2.o
+	echo 'int ext1 = 5; int *ext2 = &ext1;' | gcc -xc -c -o tmp3.o -
+	gcc -static -o tmp tmp.s tmp2.o tmp3.o
 	./tmp
 
 testx:
 	docker run --platform=linux/amd64 -v "${PWD}:/src" -w/src ccc make test
 
 clean:
-	rm -f ccc *.o *~ tmp*
+	rm -f ccc *.o *~ tmp* stage*/*
 
 .PHONY: help lint test clean process
+
+# stage0: gcc
+# stage1: ccc on host machine generated by stage0 (gcc)
+# stage2: ccc on X86-64 machine generated by stage1
+# stage3: ccc on X86-64 machine generated by stage2
+
+STAGES = stage1 stage2 stage3
+
+define PROCESS_STAGE_RULE
+$(1)/%.c: %.c
+	mkdir -p $(1)
+	gcc -E -P $$< -o $$@
+endef
+
+$(foreach stage,$(STAGES),$(eval $(call PROCESS_STAGE_RULE,$(stage))))
+
+PROCESSED_SRCS = $(foreach stage,$(STAGES),$(addprefix $(stage)/,$(SRCS)))
+
+process: $(PROCESSED_SRCS)
+
+ccc-stage1: $(OBJS)
+	$(CC) -o $@ $(OBJS) $(CFLAGS)
+
+ccc-stage2:
+	@echo TODO
+
+ccc-stage3:
+	@echo TODO

ccc.h

@@ -212,6 +212,7 @@ struct Type {
   TypeKind kind;
   bool is_typedef;     // typedef
   bool is_static;      // static
+  bool is_extern;      // extern
   bool is_incomplete;  // incomplete array
   int align;           // alignment
   Type *base;          // pointer or array

codegen.c

@@ -553,6 +553,9 @@ void emit_data(Program *prog) {
   printf(".data\n");
   for (VarList *vl = prog->globals; vl; vl = vl->next) {
     Var *var = vl->var;
+    if (var->ty->is_extern) {
+      continue;
+    }
     printf("%s:\n", var->name);
     if (!var->initializer) {
       printf("  .zero %d\n", size_of(var->ty, var->tok));

parse.c

@@ -224,7 +224,8 @@ Program *program() {
 //                | "short" | "short" "int" | "int" "short"
 //                | "int"
 //                | "long" | "long" "int" | "int" "long"
-// note that "typedef" and "static" can appear anywhere in a function body.
+// note that "typedef", "static", and "extern" can appear anywhere in a function
+// body.
 Type *type_specifier() {
   if (!is_typename(token)) {
     error_tok(token, "typename expected");
@@ -244,6 +245,7 @@ Type *type_specifier() {
 
   bool is_typedef = false;
   bool is_static = false;
+  bool is_extern = false;
 
   for (;;) {
     // read one token at a time
@@ -252,6 +254,8 @@ Type *type_specifier() {
       is_typedef = true;
     } else if (consume("static")) {
       is_static = true;
+    } else if (consume("extern")) {
+      is_extern = true;
     } else if (consume("void")) {
       base_type += VOID;
     } else if (consume("_Bool")) {
@@ -318,6 +322,7 @@ Type *type_specifier() {
   }
   ty->is_typedef = is_typedef;
   ty->is_static = is_static;
+  ty->is_extern = is_extern;
   return ty;
 }
 
@@ -772,10 +777,12 @@ Initializer *gvar_initializer(Initializer *cur, Type *ty) {
 // = type-specifier declarator type-suffix ("=" gvar-initializer)? ";"
 void global_var() {
   Type *ty = type_specifier();
+  bool is_extern = ty->is_extern;
   char *name = NULL;
   Token *tok = token;
   ty = declarator(ty, &name);
   ty = type_suffix(ty);
+  ty->is_extern = is_extern;
 
   Var *var = push_var(name, ty, false, tok);
   push_scope(name)->var = var;
@@ -989,7 +996,8 @@ Node *read_expr_stmt() {
 bool is_typename() {
   return peek("typedef") || peek("void") || peek("_Bool") || peek("char") ||
          peek("short") || peek("int") || peek("long") || peek("struct") ||
-         peek("enum") || peek("static") || find_typedef(token);
+         peek("enum") || peek("static") || peek("extern") ||
+         find_typedef(token);
 }
 
 // stmt = "return" expr ";"

tests

@@ -93,6 +93,9 @@ void voidfn(void) {}
 
 static int static_fn(void) {}
 
+extern int ext1;
+extern int *ext2;
+
 int main() {
   assert(8, ({
            int a = 3;
@@ -951,6 +954,9 @@ int main() {
   assert(3, g23, "g23");
   assert(3, *g24, "*g24");
 
+  assert(5, ext1, "ext1");
+  assert(5, *ext2, "*ext2");
+
   printf("OK\n");
   return 0;
 }

tokenize.c

@@ -161,11 +161,11 @@ bool is_alnum(char c) { return is_alpha(c) || ('0' <= c && c <= '9'); }
 
 char *starts_with_reserved(char *p) {
   // Keyword
-  static char *kw[] = {"return", "if",     "else",     "while",  "for",
-                       "char",   "int",    "sizeof",   "struct", "typedef",
-                       "short",  "long",   "void",     "_Bool",  "enum",
-                       "static", "break",  "continue", "goto",   "switch",
-                       "case",   "default"};
+  static char *kw[] = {"return", "if",      "else",     "while",  "for",
+                       "char",   "int",     "sizeof",   "struct", "typedef",
+                       "short",  "long",    "void",     "_Bool",  "enum",
+                       "static", "break",   "continue", "goto",   "switch",
+                       "case",   "default", "extern"};
   for (int i = 0; i < sizeof(kw) / sizeof(*kw); i++) {
     int len = strlen(kw[i]);
     if (startswith(p, kw[i]) && !is_alnum(p[len])) {