full build of tcc with itself - doesn't match gcc :o

11 files changed, 250 insertions, 44 deletions

diff --git a/05/Makefile b/05/Makefile
index 628d286..902a7bf 100644
--- a/05/Makefile
+++ b/05/Makefile
@@ -1,6 +1,8 @@
 TCCDIR=tcc-0.9.27
-TCC=$(TCCDIR)/tcc
-all: out04 a.out test.out README.html
+TCC0=$(TCCDIR)/tcc0
+TCC1=$(TCCDIR)/tcc1
+TCCINST=/usr/local/lib/tcc-bootstrap
+all: out04 a.out test.out README.html $(TCCDIR)/lib/libtcc1.a
 in04: *.b ../04a/out04
 	../04a/out04 main.b in04
 out04: in04 ../04/out03
@@ -11,13 +13,25 @@ out04: in04 ../04/out03
 	./out04 $< $@
 a.out: main.c *.h out04
 	./out04
-test.out: $(TCC) test.s.o test.c.o
-	$(TCC) -static -nostdlib test.s.o test.c.o -o test.out
-test.s.o: $(TCC) test.s
-	$(TCC) -static -nostdlib -c test.s -o test.s.o
-test.c.o: $(TCC) test.c
-	$(TCC) -static -nostdlib -c test.c -o test.c.o
-$(TCC): $(TCCDIR)/*.c $(TCCDIR)/*.h out04
-	cd $(TCCDIR) && ../out04 tcc.c tcc
+test.out: $(TCC0) test.s.o test.c.o
+	$(TCC0) -static -nostdlib test.s.o test.c.o -o test.out
+test.s.o: $(TCC0) test.s
+	$(TCC0) -static -nostdlib -c test.s -o test.s.o
+test.c.o: $(TCC0) test.c
+	$(TCC0) -static -nostdlib -c test.c -o test.c.o
+$(TCC0): $(TCCDIR)/*.c $(TCCDIR)/*.h out04
+	cd $(TCCDIR) && ../out04 tcc.c tcc0
+$(TCCDIR)/lib/libtcc1.a: $(TCC0) $(TCCDIR)/lib/*.[cS]
+	$(TCC0) -c $(TCCDIR)/lib/alloca86_64-bt.S -o $(TCCDIR)/lib/alloca86_64-bt.o
+	$(TCC0) -c $(TCCDIR)/lib/alloca86_64.S -o $(TCCDIR)/lib/alloca86_64.o
+	$(TCC0) -c $(TCCDIR)/lib/va_list.c -o $(TCCDIR)/lib/va_list.o
+	$(TCC0) -c $(TCCDIR)/lib/libtcc1.c -o $(TCCDIR)/lib/libtcc1.o
+	$(TCC0) -ar $(TCCDIR)/lib/libtcc1.a $(TCCDIR)/lib/*.o
+install-tcc0: $(TCCDIR)/lib/libtcc1.a $(TCCDIR)/include/*.h
+	mkdir -p $(TCCINST)/include
+	cp -r $(TCCDIR)/include/*.h $(TCCINST)/include/
+	cp -r $(TCCDIR)/lib/libtcc1.a $(TCCINST)/
+$(TCC1): $(TCC0) $(TCCINST)/libtcc1.a
+	cd $(TCCDIR) && ./tcc0 tcc.c -o tcc1
 clean:
-	rm -f out* README.html *.out *.o $(TCC)
+	rm -f out* README.html *.out *.o $(TCC0) $(TCC1) $(TCCDIR)/lib/*.[oa]
diff --git a/05/README.md b/05/README.md
index 5e5ba1d..5f54a9a 100644
--- a/05/README.md
+++ b/05/README.md
@@ -9,8 +9,8 @@ make
 ```
 
 to build our C compiler and TCC. This will take some time (approx. 25 seconds on my computer).
-A test program, `test.out` will be compiled using `tcc`. If you run
-it, you should get the output
+Two test programs will be produced: `a.out`, compiled using our C compiler, and
+`test.out`, compiled using `tcc`. If you run either one, you should get the output
 
 ```
 Hello, world!
@@ -107,8 +107,149 @@ Tokens are one of:
 - A character literal (e.g. `'a'`, `'\n'`)
 - A floating-point literal (e.g. `3.6`, `5e10`)
 
+Next, an internal representation of the program is constructed in memory.
+This is where we read the tokens `if` `(` `a` `)` `printf` `(` `"Hello!\n"` `)` `;`
+and interpret it as an if statement, whose condition is the variable `a`, and whose
+body consists of the single statement calling the `printf` function with the argument `"Hello!\n"`.
+
+Finally, we output the code for every function.
+
+## executable format
+
+This compiler's executables are much more sophisticated than the previous ones'.
+Instead of storing code and data all in one segment, we have three segments: one
+6MB segment for code (the program's functions are only allowed to use up 4MB of that, though),
+one 4MB segment for read-only data (strings), and one 4MB segment for read-write data.
+
+Well, it *should* only be read-write, but unfortunately it also has to be executable...
+
+## syscalls
+
+Of course, we need some way of making system calls in C.
+We do this with a macro, `__syscall`, which you'll find in `stdc_common.h`:
+
+```
+static unsigned char __syscall_data[] = {
+	// mov rax, [rsp+24]
+	0x48, 0x8b, 0x84, 0x24, 24, 0, 0, 0,
+	// mov rdi, rax
+	0x48, 0x89, 0xc7,
+	// mov rax, [rsp+32]
+	0x48, 0x8b, 0x84, 0x24, 32, 0, 0, 0,
+	// mov rsi, rax
+	0x48, 0x89, 0xc6,
+	// mov rax, [rsp+40]
+	0x48, 0x8b, 0x84, 0x24, 40, 0, 0, 0,
+	// mov rdx, rax
+	0x48, 0x89, 0xc2,
+	// mov rax, [rsp+48]
+	0x48, 0x8b, 0x84, 0x24, 48, 0, 0, 0,
+	// mov r10, rax
+	0x49, 0x89, 0xc2,
+	// mov rax, [rsp+56]
+	0x48, 0x8b, 0x84, 0x24, 56, 0, 0, 0,
+	// mov r8, rax
+	0x49, 0x89, 0xc0,
+	// mov rax, [rsp+64]
+	0x48, 0x8b, 0x84, 0x24, 64, 0, 0, 0,
+	// mov r9, rax
+	0x49, 0x89, 0xc1,
+	// mov rax, [rsp+16]
+	0x48, 0x8b, 0x84, 0x24, 16, 0, 0, 0,
+	// syscall
+	0x0f, 0x05,
+	// mov [rsp+8], rax
+	0x48, 0x89, 0x84, 0x24, 8, 0, 0, 0,
+	// ret
+	0xc3
+};
+
+#define __syscall(no, arg1, arg2, arg3, arg4, arg5, arg6)\
+	(((unsigned long (*)(unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long, unsigned long))__syscall_data)\
+		(no, arg1, arg2, arg3, arg4, arg5, arg6))
+```
+
+The `__syscall_data` array contains machine language instructions which perform a system call, and the
+`__syscall` macro "calls" the array as if it were a function. This is why we need a read-write-executable data
+segment -- otherwise we'd need to implement system calls in the compiler.
+
+## C standard library
+
+The C89 standard specifies a bunch of "standard library" functions which any implementation has to make available, e.g.
+`printf()`, `atoi()`, `exit()`.
+Fortunately, we don't have to write these functions in the 04 language; we can write them in C.
+
+To use a particular function, a C program needs to include the appropriate header file, e.g.
+`#include <stdio.h>` lets you use `printf()` and other I/O-related functions. Normally,
+these header files just declare what types the parameters to the functions should be,
+but we actually put the function implementations there.
+
+Let's take a look at the contents of `ctype.h`, which provides the functions `islower`, `isupper`, etc.:
+```
+#ifndef _CTYPE_H
+#define _CTYPE_H
+
+#include <stdc_common.h>
+
+int islower(int c) {
+	return c >= 'a' && c <= 'z';
+}
+
+int isupper(int c) {
+	return c >= 'A' && c <= 'Z';
+}
+
+int isalpha(int c) {
+	return isupper(c) || islower(c);
+}
+
+int isalnum(int c) {
+	return isalpha(c) || isdigit(c);
+}
+
+...
+
+#endif
+```
+The first two lines and last line prevent problems when the file is included multiple times.
+We begin by including `stdc_common.h`, which has a bunch of functions and type definitions which all
+our header files use, and then we define each of the necessary C standard library functions.
+
+
 ## limitations
 
+There are various minor ways in which this compiler doesn't actually handle all of C89.
+Here is a list of things we do wrong (this list is probably missing things, though):
+
+- [trigraphs](https://en.wikipedia.org/wiki/Digraphs_and_trigraphs#C) are not handled
+- `char[]` string literal initializers can't contain null characters (e.g. `char x[] = "a\0b";` doesn't work)
+- you can only access members of l-values (e.g. `int x = function_which_returns_struct().member` doesn't work)
+- no default-int (this is a legacy feature of C, e.g. `main() { }` can technically stand in for `int main() {}`)
+- the keyword `auto` is not handled (again, a legacy feature of C)
+- `default:` must be the last label in a switch statement.
+- external variable declarations are ignored (e.g. `extern int x; int main() { return x; }  int x = 5; ` doesn't work)
+- `typedef`s, and `struct`/`union`/`enum` declarations aren't allowed inside functions
+- conditional expressions aren't allowed inside `case` (horribly, `switch (x) { case 5 ? 6 : 3: ; }` is legal C).
+- bit-fields aren't handled
+- Technically, `1[array]` is equivalent to `array[1]`, but we don't handle that.
+- C89 has *very* weird typing rules about `void*`/`non-void*` inside conditional expressions. We don't handle that properly.
+- C89 allows calling functions without declaring them, for legacy reasons. We don't handle that.
+- Floating-point constant expressions are very limited. Only `double` literals and 0 are supported (it was hard enough
+to parse floating-point literals in a language without floating-point variables!)
+- Floating-point literals can't have their integer part greater than 2<sup>64</sup>-1.
+- Redefining a macro is always an error, even if it's the same definition.
+- You can't have a variable/function/etc. called `defined`.
+- Various little things about when macros are evaluated in some contexts.
+setjmp.h:// @NONSTANDARD: we don't actually support setjmp
+stddef.h:// @NONSTANDARD: we don't have wchar_t
+stdlib.h:// @NONSTANDARD: we don't define MB_CUR_MAX or any of the mbtowc functions
+time.h:// @NONSTANDARD(except in UTC+0): we don't support local time in timezones other than UTC+0.
+time.h: // @NONSTANDARD-ish.
+
+
+Also, the keywords `signed`, `volatile`, `register`, and `const` are all ignored. This shouldn't have an effect
+on any legal C program, though.
+
 ## modifications of tcc's source code
 
 
diff --git a/05/main.c b/05/main.c
index e669aec..8d91650 100644
--- a/05/main.c
+++ b/05/main.c
@@ -1,26 +1,15 @@
-/* #define _STDLIB_DEBUG */
-/* #include <math.h> */
 #include <stdio.h>
-/* #include <signal.h> */
-/* #include <stdlib.h> */
-/* #include <string.h> */
-/* #include <time.h> */
-/* #include <float.h> */
-/* #include <setjmp.h> */
-/*  */
+#include <stdlib.h>
 
 int main(int argc, char **argv) {
-	int *p = 0x100;
-	p += 1;
-	switch (5) {
-	case 5:
-		switch (6) {
-		default:;
-		}
-	case 6:
-		;
-	}
-	printf("%p\n",p);
+	printf("%p\n", malloc(1024*16));
+	int *list = malloc(1024*4);
+	printf("%p \n",list);
+	list[1023] = 77;
+	list = realloc(list, 1024*64);
+	printf("%p \n",list);
+	printf("%d\n",list[1023]);
+	free(list);
 	return 0;
 }
 
diff --git a/05/stdlib.h b/05/stdlib.h
index b93d4aa..cbfd0b5 100644
--- a/05/stdlib.h
+++ b/05/stdlib.h
@@ -73,12 +73,25 @@ void *realloc(void *ptr, size_t size) {
 		free(ptr);
 		return NULL;
 	}
+#if 0
+	// this (better) implementation doesn't seem to be copying stuff to the
+	// new mapping properly
 	uint64_t *memory = (char *)ptr - 16;
 	uint64_t old_size = *memory;
 	uint64_t *new_memory = _mremap(memory, old_size, size, MREMAP_MAYMOVE);
 	if ((uint64_t)new_memory > 0xffffffffffff0000) return NULL;
 	*new_memory = size;
 	return (char *)new_memory + 16;
+#endif
+
+	uint64_t *memory = (char *)ptr - 16;
+	uint64_t old_size = *memory;
+	void *new = malloc(size);
+	char *new_dat = (char *)new + 16;
+	*(uint64_t *)new = size;
+	memcpy(new_dat, ptr, old_size);
+	free(ptr);
+	return new_dat;
 }
 
 
diff --git a/05/tcc-0.9.27/.gitignore b/05/tcc-0.9.27/.gitignore
index a50ca5e..f7853f0 100644
--- a/05/tcc-0.9.27/.gitignore
+++ b/05/tcc-0.9.27/.gitignore
@@ -16,6 +16,8 @@
 a.out
 tcc_g
 tcc
+tcc[0123456789]
+tcc[0123456789]a
 *-tcc
 libtcc*.def
 
diff --git a/05/tcc-0.9.27/config.h b/05/tcc-0.9.27/config.h
index d363b97..95ec14d 100644
--- a/05/tcc-0.9.27/config.h
+++ b/05/tcc-0.9.27/config.h
@@ -1,8 +1,10 @@
 #define TCC_VERSION "0.9.27"
 #define CONFIG_TCC_STATIC 1
-#define CONFIG_TCC_ELFINTERP "/XXX"
-#define CONFIG_TCC_CRT_PREFIX "/XXX"
-#define CONFIG_SYSROOT "/XXX"
+//#define CONFIG_TCC_ELFINTERP "/XXX"
+//#define CONFIG_TCC_CRT_PREFIX "/XXX"
+//#define CONFIG_SYSROOT "/XXX"
 #define inline
 #define TCC_TARGET_X86_64 1
 #define ONE_SOURCE 1
+#define CONFIG_LDDIR "lib/x86_64-linux-gnu"
+#define CONFIG_TCCDIR "/usr/local/lib/tcc-bootstrap"
diff --git a/05/tcc-0.9.27/lib/libtcc1.c b/05/tcc-0.9.27/lib/libtcc1.c
index 0e46618..633696b 100644
--- a/05/tcc-0.9.27/lib/libtcc1.c
+++ b/05/tcc-0.9.27/lib/libtcc1.c
@@ -613,10 +613,17 @@ unsigned long long __fixunsxfdi (long double a1)
         return 0;
 }
 
+
+static long double negate_ld(long double d) {
+	register unsigned long long *p = (unsigned long long *)&d;
+	p[1] ^= 1ul<<15;
+	return *(long double *)p;
+}
+
 long long __fixxfdi (long double a1)
 {
     long long ret; int s;
-    ret = __fixunsxfdi((s = a1 >= 0) ? a1 : -a1);
+    ret = __fixunsxfdi((s = a1 >= 0) ? a1 : negate_ld(a1));
     return s ? ret : -ret;
 }
 #endif /* !ARM */
diff --git a/05/tcc-0.9.27/stdlib.h b/05/tcc-0.9.27/stdlib.h
index b93d4aa..cbfd0b5 100644
--- a/05/tcc-0.9.27/stdlib.h
+++ b/05/tcc-0.9.27/stdlib.h
@@ -73,12 +73,25 @@ void *realloc(void *ptr, size_t size) {
 		free(ptr);
 		return NULL;
 	}
+#if 0
+	// this (better) implementation doesn't seem to be copying stuff to the
+	// new mapping properly
 	uint64_t *memory = (char *)ptr - 16;
 	uint64_t old_size = *memory;
 	uint64_t *new_memory = _mremap(memory, old_size, size, MREMAP_MAYMOVE);
 	if ((uint64_t)new_memory > 0xffffffffffff0000) return NULL;
 	*new_memory = size;
 	return (char *)new_memory + 16;
+#endif
+
+	uint64_t *memory = (char *)ptr - 16;
+	uint64_t old_size = *memory;
+	void *new = malloc(size);
+	char *new_dat = (char *)new + 16;
+	*(uint64_t *)new = size;
+	memcpy(new_dat, ptr, old_size);
+	free(ptr);
+	return new_dat;
 }
 
 
diff --git a/05/tcc-0.9.27/tcc.h b/05/tcc-0.9.27/tcc.h
index c7f6c72..6c3c10c 100644
--- a/05/tcc-0.9.27/tcc.h
+++ b/05/tcc-0.9.27/tcc.h
@@ -30,14 +30,15 @@
 #include <string.h>
 #include <errno.h>
 #include <math.h>
-#ifdef __GNUC__
+#include <float.h>
+#if defined __GNUC__ || defined __TINYC__
 #include <fcntl.h>
 #endif
 #include <setjmp.h>
 #include <time.h>
 
 #ifndef _WIN32
-#ifdef __GNUC__
+#if defined __GNUC__ || defined __TINYC__
 # include <unistd.h>
 # include <sys/time.h>
 #endif
@@ -211,7 +212,8 @@ extern long double strtold (const char *__nptr, char **__endptr);
 #  define CONFIG_TCC_SYSINCLUDEPATHS \
         "{B}/include" \
     ":" ALSO_TRIPLET(CONFIG_SYSROOT "/usr/local/include") \
-    ":" ALSO_TRIPLET(CONFIG_SYSROOT "/usr/include")
+    ":" ALSO_TRIPLET(CONFIG_SYSROOT "/usr/include") \
+    ":/usr/include/x86_64-linux-gnu"
 # endif
 #endif
 
diff --git a/05/tcc-0.9.27/tccgen.c b/05/tcc-0.9.27/tccgen.c
index 7f5d1e6..ecb7812 100644
--- a/05/tcc-0.9.27/tccgen.c
+++ b/05/tcc-0.9.27/tccgen.c
@@ -2458,6 +2458,16 @@ static void gen_cast_s(int t)
     gen_cast(&type);
 }
 
+static long double negate_ld(long double d) {
+	#if LDBL_MANT_DIG == 64
+	register unsigned long long *p = (unsigned long long *)&d;
+	p[1] ^= 1ul<<15;
+	return *(long double *)p;
+	#else
+	return -d;
+	#endif 
+}
+
 static void gen_cast(CType *type)
 {
     int sbt, dbt, sf, df, c, p;
@@ -2499,12 +2509,12 @@ static void gen_cast(CType *type)
                     if ((sbt & VT_UNSIGNED) || !(vtop->c.i >> 63))
                         vtop->c.ld = vtop->c.i;
                     else
-                        vtop->c.ld = -(long double)-vtop->c.i;
+                        vtop->c.ld = negate_ld((long double)-vtop->c.i);
                 } else if(!sf) {
                     if ((sbt & VT_UNSIGNED) || !(vtop->c.i >> 31))
                         vtop->c.ld = (uint32_t)vtop->c.i;
                     else
-                        vtop->c.ld = -(long double)-(uint32_t)vtop->c.i;
+                        vtop->c.ld = negate_ld((long double)-(uint32_t)vtop->c.i);
                 }
 
                 if (dbt == VT_FLOAT)
diff --git a/markdown.c b/markdown.c
index 27a4a1b..3b812aa 100644
--- a/markdown.c
+++ b/markdown.c
@@ -19,7 +19,12 @@ static void output_md_text(FILE *out, int *flags, int line_number, const char *t
 
 	for (p = text; *p; ++p) {
 		if ((*flags & FLAG_CODE) && *p != '`') {
-			putc(*p, out);
+			switch (*p) {
+			case '<': fprintf(out, "&lt;"); break;
+			case '>': fprintf(out, "&gt;"); break;
+			case '&': fprintf(out, "&amp;"); break;
+			default: putc(*p, out); break;
+			}
 			continue;
 		}
 		switch (*p) {
@@ -198,10 +203,18 @@ int main(int argc, char **argv) {
 			fprintf(out, "<pre><code>\n");
 			
 			while (fgets(line, sizeof line, in)) {
+				char *p;
 				++line_number;
 				if (strncmp(line, "```", 3) == 0)
 					break;
-				fprintf(out, "%s", line);
+				for (p = line; *p; ++p) {
+					switch (*p) {
+					case '<': fprintf(out, "&lt;"); break;
+					case '>': fprintf(out, "&gt;"); break;
+					case '&': fprintf(out, "&amp;"); break;
+					default: fputc(*p, out); break;
+					}
+				}
 			}
 
 			fprintf(out, "</code></pre>\n");