meta: add LTO support and enable it by default

* Set -flto
* Use gcc-ar to build the archive so the LTO plugin is loaded and LTO
  sections aren't stripped off the object files
* Slightly conservative KEEP()s in the linker script (not required but
  I'm paranoid)
* Make the drivers __attribute__((externall_visible)) so they are
  generated by the link-time back-end and not left in the archive
* Remove the unused isappli/optnum parameters of main, which have been
  unused for years, are irrelevant on fx-CG and lead to a link-time
  warning with LTO. I'll add APIs to access them later.

CMakeLists.txt

@@ -259,7 +259,7 @@ fxconv_declare_assets(${ASSETS_FX} ${ASSETS_CG})
 include_directories(
   "${PROJECT_SOURCE_DIR}/include"
   "${PROJECT_BINARY_DIR}/include")
-add_compile_options(-Wall -Wextra -std=c11 -Os -g -fstrict-volatile-bitfields -mtas)
+add_compile_options(-Wall -Wextra -std=c11 -Os -g -fstrict-volatile-bitfields -mtas -flto)
 
 if("${FXSDK_PLATFORM_LONG}" STREQUAL fx9860G)
   add_compile_definitions(FX9860G)
@@ -297,6 +297,12 @@ endif()
 
 set_target_properties("${NAME}" PROPERTIES OUTPUT_NAME "${NAME}")
 
+# Generate the archive with gcc-ar instead of ar as it will load the LTO plugin
+# which is required to generate a usable archive.
+set(CMAKE_C_ARCHIVE_CREATE "${CMAKE_C_COMPILER_AR} qcs <TARGET> <OBJECTS>")
+# Also the ranlib rule (useless because ar is passed the s flag anyway)
+set(CMAKE_C_ARCHIVE_FINISH "${CMAKE_C_COMPILER_RANLIB} <TARGET>")
+
 # Generate linker scripts
 macro(generate_linker_script OUTPUT INPUT OPTIONS)
   add_custom_command(

fx9860g.ld.c

@@ -61,19 +61,19 @@ SECTIONS
 		*(.text.entry)
 
 		_bctors = . ;
-		*(.ctors .ctors.*)
+		KEEP(*(.ctors .ctors.*))
 		_ectors = . ;
 
 		_bdtors = . ;
-		*(.dtors .dtors.*)
+		KEEP(*(.dtors .dtors.*))
 		_edtors = . ;
 
 		_gint_exch_start = . ;
-		*(.gint.exch)
+		KEEP(*(.gint.exch))
 		_gint_exch_size = ABSOLUTE(. - _gint_exch_start);
 
 		_gint_tlbh_start = . ;
-		*(.gint.tlbh)
+		KEEP(*(.gint.tlbh))
 		_gint_tlbh_size = ABSOLUTE(. - _gint_tlbh_start);
 
 		*(.text .text.*)

fxcg50.ld.c

@@ -50,19 +50,19 @@ SECTIONS
 		*(.text.entry)
 
 		_bctors = . ;
-		*(.ctors .ctors.*)
+		KEEP(*(.ctors .ctors.*))
 		_ectors = . ;
 
 		_bdtors = . ;
-		*(.dtors .dtors.*)
+		KEEP(*(.dtors .dtors.*))
 		_edtors = . ;
 
 		_gint_exch_start = . ;
-		*(.gint.exch)
+		KEEP(*(.gint.exch))
 		_gint_exch_size = ABSOLUTE(. - _gint_exch_start);
 
 		_gint_tlbh_start = . ;
-		*(.gint.tlbh)
+		KEEP(*(.gint.tlbh))
 		_gint_tlbh_size = ABSOLUTE(. - _gint_tlbh_start);
 
 		*(.text .text.*)

include/gint/defs/attributes.h

@@ -34,6 +34,9 @@
 /* Transparent unions */
 #define GTRANSPARENT	__attribute__((transparent_union))
 
+/* Functions and globals that are visible through whole-program optimization */
+#define GVISIBLE  __attribute__((externally_visible))
+
 /* Weak symbols */
 #define GWEAK		__attribute__((weak))
 

include/gint/drivers.h

@@ -267,7 +267,7 @@ typedef void **gint_world_t;
    the section name and the linker then sorts by name. If your driver has a
    level lower than 10, you must add a leading 0. */
 #define GINT_DECLARE_DRIVER(level, name) \
-	GSECTION(".gint.drivers." #level) extern gint_driver_t name;
+	GSECTION(".gint.drivers." #level) GVISIBLE extern gint_driver_t name;
 
 //---
 // Internal driver control

src/kernel/start.c

@@ -34,7 +34,7 @@ extern void (*bctors)(void), (*ectors)(void);
 extern void (*bdtors)(void), (*edtors)(void);
 
 /* User-provided main() function */
-int main(int isappli, int optnum);
+int main(void);
 
 /* Whether to restart main through the OS menu rather than returning */
 int8_t gint_restart = 0;
@@ -188,7 +188,10 @@ static int start2(int isappli, int optnum)
 	   what it wants in exit() after main() finishes executing */
 	if(!setjmp(gint_exitbuf)) {
 		callarray(&bctors, &ectors);
-		exit(main(isappli, optnum));
+		// TODO: record isappli and optnum in globals
+		(void)isappli;
+		(void)optnum;
+		exit(main());
 	}
 	else {
 		callarray(&bdtors, &edtors);