-Werror is causing some heart ache

	Hardware:

	- EST SBC8260
	- MPC8260 Rev A.1 CPU/CPM/BUS: 200/133/66 Mhz

	Environment:

	- Linux: 2.4.0-test3

In one of the recent changesets, the CFLAGS definition in ./Makefile
was modified to add -Werror.  The -Werror option to gcc causes it to
treat warnings are errors.

In order to get arch/ppc/kernel/setup.c to compile, a functional
prototype for m8260_init needed to be created.

dell 384} bk diffs -u -r1.40 arch/ppc/kernel/setup.c
===== arch/ppc/kernel/setup.c 1.40 vs 1.41 =====

--- 1.40/arch/ppc/kernel/setup.c        Sun Jul  9 22:42:49 2000
+++ 1.41/arch/ppc/kernel/setup.c        Mon Jul 10 16:37:44 2000

@@ -63,6 +63,12 @@
                     unsigned long r6,
                     unsigned long r7);

+extern void m8260_init(unsigned long r3,
+                      unsigned long r4,
+                      unsigned long r5,
+                      unsigned long r6,
+                      unsigned long r7);
+
 extern void apus_init(unsigned long r3,
                       unsigned long r4,
                       unsigned long r5,

I am struggling to get arch/ppc/kernel/m8260_setup.c to compile.

gmake -C  arch/ppc/kernel
gmake[1]: Entering directory `/usr/people/diekema/bk/linux-2.3/arch/ppc/kernel'
/opt/hardhat/devkit/ppc/8xx/bin/powerpc-linux-gcc -D__KERNEL__ -I/usr/people/diekema/bk/linux-2.3/include -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer -Werror -fno-strict-aliasing -D__powerpc__ -fsigned-char -msoft-float -pipe -ffixed-r2 -Wno-uninitialized -mmultiple -mstring   -c -o m8260_setup.o m8260_setup.c
cc1: warnings being treated as errors
m8260_setup.c: In function `abort':
m8260_setup.c:105: warning: `noreturn' function does return
gmake[1]: *** [m8260_setup.o] Error 1
gmake[1]: Leaving directory `/usr/people/diekema/bk/linux-2.3/arch/ppc/kernel'
gmake: *** [_dir_arch/ppc/kernel] Error 2

If I used the -fno-builtin option, then gcc will compile this file.  I
added -fno-builtin to CFLAGS, and ran into problems with
arch/ppc/mm/init.c.

gmake[2]: Entering directory `/usr/people/diekema/bk/linux-2.3/arch/ppc/mm'
/opt/hardhat/devkit/ppc/8xx/bin/powerpc-linux-gcc -D__KERNEL__ -I/usr/people/diekema/bk/linux-2.3/include -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer -Werror -fno-builtin -fno-strict-aliasing -D__powerpc__ -fsigned-char -msoft-float -pipe -ffixed-r2 -Wno-uninitialized -mmultiple -mstring   -c -o init.o init.c
cc1: warnings being treated as errors
init.c:614: warning: `get_mem_prop' declared `static' but never defined
gmake[2]: *** [init.o] Error 1
gmake[2]: Leaving directory `/usr/people/diekema/bk/linux-2.3/arch/ppc/mm'
gmake[1]: *** [first_rule] Error 2
gmake[1]: Leaving directory `/usr/people/diekema/bk/linux-2.3/arch/ppc/mm'
gmake: *** [_dir_arch/ppc/mm] Error 2

It will take a fair amount of time to clean up all the warnings...

I am going to remove -fno-builtin that I just added to to CFLAGS and -Werror.
I want to to see if I can compile Linux like I use to.  Yup, this did the trick.

dell 533} bk diffs -u -r1.62 Makefile
===== Makefile 1.62 vs 1.63 =====

--- 1.62/Makefile       Sun Jul  9 22:42:49 2000
+++ 1.63/Makefile       Mon Jul 10 19:09:55 2000

@@ -84,7 +84,7 @@

 CPPFLAGS := -D__KERNEL__ -I$(HPATH)

-CFLAGS := $(CPPFLAGS) -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer -Werror
+CFLAGS := $(CPPFLAGS) -Wall -Wstrict-prototypes -O2 -fomit-frame-pointer
 AFLAGS := -D__ASSEMBLY__ $(CPPFLAGS)

 # use '-fno-strict-aliasing', but only if the compiler can take it

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

diekema_jon wrote:

In one of the recent changesets, the CFLAGS definition in ./Makefile
was modified to add -Werror.

That's been one of the many discussions about -test3 lately.   Depending
upon the time of day you picked up the sources, there are lots of
other things likely to be broken.


	-- Dan

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Hello!

In one of the recent changesets, the CFLAGS definition in ./Makefile
was modified to add -Werror.  The -Werror option to gcc causes it to
treat warnings are errors.

I have already complained about this to the main linux list. Warnings are
bad, but at this stage of the game it is an ordinary user recompiling the
kernel who suffers, not the sloppy developer.

Linux moves to the next stable release. What if in some configurations
some variables are defined but not used? Should we add more and more
ifdef's to fix those subtle warnings (and possibly break other
configurations without leaving the time to fix them)? I don't think so.

Regards,
Pavel Roskin


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/