fix -fno-plt failing

Makefile

@@ -7,11 +7,14 @@ TESTDIR:= test
 BITS ?= $(shell getconf LONG_BIT)
 
 # -mpreferred-stack-boundary=3 messes up the stack and kills SSE!
-# -fno-plt
-COPTFLAGS=-Os -fvisibility=hidden -fwhole-program -fno-plt \
-  -ffast-math -funsafe-math-optimizations -fno-stack-protector -fomit-frame-pointer \
-  -fno-exceptions -fno-unwind-tables -fno-asynchronous-unwind-tables
-CXXOPTFLAGS=$(COPTFLAGS) \
+#COPTFLAGS=-Os -fvisibility=hidden -fwhole-program -fno-plt \
+#  -ffast-math -funsafe-math-optimizations -fno-stack-protector -fomit-frame-pointer \
+#  -fno-exceptions -fno-unwind-tables -fno-asynchronous-unwind-tables
+
+COPTFLAGS=-O2 -fno-plt -fno-stack-protector -fno-stack-check -fno-unwind-tables \
+  -fno-asynchronous-unwind-tables -fomit-frame-pointer -ffast-math -no-pie \
+  -fno-pic -fno-PIE -m64 -march=core2 -ffunction-sections -fdata-sections
+CXXOPTFLAGS=$(COPTFLAGS) -fno-exceptions \
   -fno-rtti -fno-enforce-eh-specs -fnothrow-opt -fno-use-cxa-get-exception-ptr \
   -fno-implicit-templates -fno-threadsafe-statics -fno-use-cxa-atexit
 
@@ -19,26 +22,30 @@ CFLAGS=-Wall -Wextra -Wpedantic -std=gnu11 -nostartfiles -fno-PIC $(COPTFLAGS)
 CXXFLAGS=-Wall -Wextra -Wpedantic -std=c++11 $(CXXOPTFLAGS) -nostartfiles -fno-PIC
 
 ASFLAGS=-I $(SRCDIR)/
+LDFLAGS_ :=
 ifeq ($(BITS),32)
-LDFLAGS=-m elf_i386
+LDFLAGS += -m32
 ASFLAGS += -f elf32
+LDFLAGS_ := -m32
 else
-LDFLAGS=-m elf_x86_64
+LDFLAGS += -m64
 ASFLAGS += -f elf64
+LDFLAGS_ := -m64
 endif
-LDFLAGS_=$(LDFLAGS) -T $(LDDIR)/link.ld --oformat=binary
+LDFLAGS += -nostartfiles -nostdlib
+LDFLAGS_ := $(LDFLAGS_) -T $(LDDIR)/link.ld -Wl,--oformat=binary $(LDFLAGS)
 
 CFLAGS   += -m$(BITS) $(shell pkg-config --cflags sdl2)
 CXXFLAGS += -m$(BITS) $(shell pkg-config --cflags sdl2)
 
 LIBS=-lc
 
-ASFLAGS += -DUSE_INTERP -DALIGN_STACK -DUSE_DT_DEBUG -DNO_START_ARG
+ASFLAGS += -DUSE_INTERP -DALIGN_STACK #-DUSE_DT_DEBUG #-DNO_START_ARG
 
 NASM    ?= nasm
 PYTHON3 ?= python3
 
-all: $(BINDIR)/hello-crt $(BINDIR)/sdl-crt $(BINDIR)/flag $(BINDIR)/hello-_start
+all: $(BINDIR)/hello-crt $(BINDIR)/sdl-crt $(BINDIR)/flag-crt $(BINDIR)/hello-_start
 
 LIBS += $(filter-out -pthread,$(shell pkg-config --libs sdl2)) -lX11 #-lGL
 
@@ -50,13 +57,18 @@ clean:
 
 .SECONDARY:
 
+$(OBJDIR)/%.lto.o: $(SRCDIR)/%.c $(OBJDIR)/
+	$(CC) -flto $(CFLAGS) -c "$<" -o "$@"
+$(OBJDIR)/%.lto.o: $(TESTDIR)/%.c $(OBJDIR)/
+	$(CC) -flto $(CFLAGS) -c "$<" -o "$@"
+
 $(OBJDIR)/%.o: $(SRCDIR)/%.c $(OBJDIR)/
 	$(CC) $(CFLAGS) -c "$<" -o "$@"
 $(OBJDIR)/%.o: $(TESTDIR)/%.c $(OBJDIR)/
 	$(CC) $(CFLAGS) -c "$<" -o "$@"
 
-$(OBJDIR)/%.start.o: $(OBJDIR)/%.o $(OBJDIR)/crt1.o
-	$(LD) $(LDFLAGS) -r -o "$@" $^
+$(OBJDIR)/%.start.o: $(OBJDIR)/%.lto.o $(OBJDIR)/crt1.lto.o
+	$(CC) $(LDFLAGS) -r -o "$@" $^
 
 $(OBJDIR)/symbols.%.asm: $(OBJDIR)/%.o
 	$(PYTHON3) ./smol.py $(SMOLFLAGS) $(LIBS) "$<" "$@"
@@ -70,10 +82,10 @@ $(OBJDIR)/stub.%.start.o: $(OBJDIR)/symbols.%.start.asm $(SRCDIR)/header32.asm \
 	$(NASM) $(ASFLAGS) $< -o $@
 
 $(BINDIR)/%: $(OBJDIR)/%.o $(OBJDIR)/stub.%.o $(BINDIR)/
-	$(LD) -Map=$(BINDIR)/$*.map $(LDFLAGS_) $(OBJDIR)/$*.o $(OBJDIR)/stub.$*.o -o "$@"
+	$(CC) -Wl,-Map=$(BINDIR)/$*.map $(LDFLAGS_) $(OBJDIR)/$*.o $(OBJDIR)/stub.$*.o -o "$@"
 
 $(BINDIR)/%-crt: $(OBJDIR)/%.start.o $(OBJDIR)/stub.%.start.o $(BINDIR)/
-	$(LD) -Map=$(BINDIR)/$*-crt.map $(LDFLAGS_) $(OBJDIR)/$*.start.o $(OBJDIR)/stub.$*.start.o -o "$@"
+	$(CC) -Wl,-Map=$(BINDIR)/$*-crt.map $(LDFLAGS_) $(OBJDIR)/$*.start.o $(OBJDIR)/stub.$*.start.o -o "$@"
 
 .PHONY: all clean
 

src/crt1.c

@@ -2,6 +2,7 @@
 #include <stddef.h>
 
 extern int main(int argc, char* argv[]);
+
 extern int __libc_start_main(int (*main)(int, char**),
         int argc, char** argv,
         void (*init)(void), void(*fini)(void),
@@ -9,14 +10,30 @@ extern int __libc_start_main(int (*main)(int, char**),
         void* stack) __attribute__((__noreturn__));
 
 __attribute__((__externally_visible__, __section__(".text.startup._start"),
-            __noreturn__))
+            __noreturn__, __used__
+#ifdef __x86_64__
+        , __naked__
+#endif
+))
 int _start(void* stack) {
     // TODO: _dl_fini etc.
     int argc=*(size_t*)stack;
     char** argv=(void*)(&((size_t*)stack)[1]);
 
-    // TODO: fix this. it borks with -fno-plt
+// avoid problems when -fno-plt is enabled
+#ifdef __x86_64__
+    asm volatile("xor  %%ecx, %%ecx\n"
+                 "push %%rcx\n"
+                 "push %%rcx\n"
+                 "pop  %%r8\n"
+                 "pop  %%r9\n"
+                 "call *__libc_start_main@GOTPCREL(%%rip)\n"
+            :
+            :"S"(argc), "D" (main), "d" (argv)
+            :);
+#else
     __libc_start_main(main, argc, argv, NULL, NULL, NULL, (void*)stack);
+#endif
 
     __builtin_unreachable();
 }

src/loader64.asm

@@ -142,7 +142,8 @@ repne scasd ; technically, scasq should be used, but ehhhh
     mov rdi, rsp
 %endif
 %ifdef ALIGN_STACK
-   push rax
+       ; apparently not needed?
+; push rax
 %endif
 %ifdef USE_DL_FINI
    xchg rsi, r13 ; _dl_fini

test/flag.c

@@ -10,9 +10,9 @@
 
 #define COLOR(r, g, b) ((r << 16) + (g << 8) + b)
 
-__attribute__((__used__,__externally_visible__))
-void _start() {
-/*int main() {*/
+/*__attribute__((__used__,__externally_visible__))
+void _start() {*/
+int main() {
 #ifdef MAKE_ESC_WORK
   Atom wmDeleteMessage;
 #endif