===================================================================
@@ -395,6 +395,11 @@ extern GTY(()) int darwin_ms_struct;
%{Zforce_cpusubtype_ALL:-force_cpusubtype_ALL} \
%{static}"
+/* Default ASM_DEBUG_SPEC. Darwin's as cannot currently produce dwarf
+ debugging data. */
+
+#define ASM_DEBUG_SPEC "%{g*:%{!g0:%{!gdwarf*:--gstabs}}}"
+
/* We still allow output of STABS. */
#define DBX_DEBUGGING_INFO 1
===================================================================
@@ -41,6 +41,12 @@ along with GCC; see the file COPYING3. If not see
%{mdynamic-no-pic: %n'-mdynamic-no-pic' overrides '-pie', '-fpie' or '-fPIE'; \
:-pie}}"
+/* Only ask as for debug data if the debug style is stabs (since as doesn't
+ yet generate dwarf.) */
+
+#undef ASM_DEBUG_SPEC
+#define ASM_DEBUG_SPEC "%{g*:%{!g0:%{gstabs:--gstabs}}}"
+
#undef ASM_OUTPUT_ALIGNED_COMMON
#define ASM_OUTPUT_ALIGNED_COMMON(FILE, NAME, SIZE, ALIGN) \
do { \
As things stand, Darwin's 'as' cannot produce dwarf debug. Unfortunately, if "-g" is given to -x assembler-with-cpp ... this results in 'as' being called with --stabs. So, it can be the case that an exe ends up with some stabs debug in it - usually with a source file of "{standard input}" (and without an OSO entry). This upsets the OSO parser in machoread.c (in gdb) causing gdb to fail with an OSO != NULL assert. === This patch prevents "as" being asked to give debug output unless stabs is the chosen debug scheme (so for the default on Darwin >= 9 asm files will not get any debug data). It doesn't prevent the scenario completely - the user could still produce the situation manually (I have a gdb patch to work around it) ... === However, this does mean that a standard (ppc) build of gcc doesn't abort gdb (because the .S files in libgcc_s are causing this on ppc). OK for trunk? Iain