klibc - Feb 2005 - [PATCH] make DEBUG=true

udev knows about DEBUG=true and builds a debug binary. Maybe klibc
should do the same?
Was there a reason for -g in the i386 MCONFIG?

Index: MCONFIG
==================================================================--- MCONFIG
(revision 1003)
+++ MCONFIG	(working copy)
@@ -62,5 +62,11 @@
 #
 include $(KLIBSRC)/arch/$(ARCH)/MCONFIG
 
+ifeq ($(DEBUG),true)
+STRIP	= /bin/true -Since_we_are_debugging
+OPTFLAGS	+= -O1 -g
+endif
+
+
 # How to tell the linker main() is the entrypoint
 EMAIN ?= -e main
Index: klibc/arch/i386/MCONFIG
==================================================================---
klibc/arch/i386/MCONFIG	(revision 1003)
+++ klibc/arch/i386/MCONFIG	(working copy)
@@ -13,7 +13,7 @@
 
 gcc_major := $(shell echo __GNUC__ | $(CC) -E -xc - | tail -n 1)
 
-OPTFLAGS = $(REGPARM) -march=i386 -Os -g
+OPTFLAGS = $(REGPARM) -march=i386 -Os
 
 ifeq ($(gcc_major),3)
 REGPARM  := $(REGPARM_OPT)

Olaf Hering wrote:
> udev knows about DEBUG=true and builds a debug binary. Maybe klibc
> should do the same?
> Was there a reason for -g in the i386 MCONFIG?
Yes, for i386 you can strip off all debugging information, so by 
compiling with -g and then stripping we get debugging and nondebugging 
binaries.  That's not true for all architectures, in particular, it's 
not true for x86-64 or any other architecture that expects glibc to use 
libunwind (klibc doesn't, for obvious reasons.)

I really dislike DEBUG=true though, since FOO=1 (or occationally FOO=y) 
seems to be the more common idiom.

	-hpa