Re: [PATCH] Support SPARSE in Makefile, better SPARSE_FLAGS

Pavel Roskin [off-list ref] writes:

+# explicitly what architecture to check for.
+SPARSE = sparse
+SPARSE_FLAGS = -D__$(shell uname -i)__

        : siamese; uname --version
        uname (coreutils) 5.2.1
        Written by David MacKenzie.

        Copyright (C) 2004 Free Software Foundation, Inc.
        This is free software; see the source for copying conditions.
        There is NO
        warranty; not even for MERCHANTABILITY or FITNESS FOR A
        PARTICULAR PURPOSE.
        : siamese; uname -i
        Try `uname --help' for more information.

Better alternatives?

On Thu, 2005-09-29 at 22:46 -0700, Junio C Hamano wrote:

Pavel Roskin [off-list ref] writes:

quoted

+# explicitly what architecture to check for.
+SPARSE = sparse
+SPARSE_FLAGS = -D__$(shell uname -i)__

        : siamese; uname --version
        uname (coreutils) 5.2.1
        Written by David MacKenzie.

Apparently my uname 5.2.1 was heavily patched by Fedora.  Note that
"uname -m" is not good enough, as it would give us i686 when i386 is
really needed.

        Copyright (C) 2004 Free Software Foundation, Inc.
        This is free software; see the source for copying conditions.
        There is NO
        warranty; not even for MERCHANTABILITY or FITNESS FOR A
        PARTICULAR PURPOSE.
        : siamese; uname -i
        Try `uname --help' for more information.

Better alternatives?

I hate to say that, but a better alternative is to fix sparse to act
like the native compiler by default (possibly with options to imitate
other architectures or to be fully arch-neutral).

I have some hacks in mind, but I don't really like them:

Not good for i686 with unpatched uname:
SPARSE_FLAGS = __$(shell uname -i 2>/dev/null || uname -m)__

gcc specific:
SPARSE_FLAGS = __$(shell $CC -dumpmachine | sed 's/-.*//')__

bash specific:
SPARSE_FLAGS = __$(shell echo $$HOSTTYPE)__

In any case, having SPARSE variable would be convenient.

-- 
Regards,
Pavel Roskin

Pavel Roskin wrote:

I hate to say that, but a better alternative is to fix sparse to act
like the native compiler by default (possibly with options to imitate
other architectures or to be fully arch-neutral).

$(CC) $(CFLAGS) -E -dM -x c /dev/null -o builtin.h

... will output a file containing all the buildin macros that you can 
feed to sparse with -include.

Replacing -x c with, say, -x c++ gives you the builtins for C++; 
including $(CFLAGS) gives the appropriate set of macros for any 
particular combination of options (which can affect the builtin macro set.)

	-hpa

On Fri, 2005-09-30 at 12:19 -0700, H. Peter Anvin wrote:

Pavel Roskin wrote:

quoted

I hate to say that, but a better alternative is to fix sparse to act
like the native compiler by default (possibly with options to imitate
other architectures or to be fully arch-neutral).

$(CC) $(CFLAGS) -E -dM -x c /dev/null -o builtin.h

... will output a file containing all the buildin macros that you can 
feed to sparse with -include.

I know.  That's what I'm using in the wrapper (plus -m64 and some
warnings).  But it should be the default.  Until then, hassle-free
sparse support in the Makefile is only possible for the projects that
already know the architecture (e.g. the Linux kernel).

-- 
Regards,
Pavel Roskin

Pavel Roskin wrote:

I know.  That's what I'm using in the wrapper (plus -m64 and some
warnings).  But it should be the default.  Until then, hassle-free
sparse support in the Makefile is only possible for the projects that
already know the architecture (e.g. the Linux kernel).

I think that's debatable.  It introduces main-compiler dependencies into 
sparse which is undesirable.

A much simpler option would be to write a "sparsegcc" script which would 
be invoked just like gcc, extract the appropriate macro information 
based on options, and then invoke sparse.

	-hpa

Quoting "H. Peter Anvin" [off-list ref]:

Pavel Roskin wrote:

quoted

I know.  That's what I'm using in the wrapper (plus -m64 and some
warnings).  But it should be the default.  Until then, hassle-free
sparse support in the Makefile is only possible for the projects that
already know the architecture (e.g. the Linux kernel).

I think that's debatable.  It introduces main-compiler dependencies into
sparse which is undesirable.

I see sparse is already moving in this direction.  Right now, it's somewhere in
the middle, which is quite inconvenient.  It hardcodes gcc version numbers but
not the architecture.

A much simpler option would be to write a "sparsegcc" script which would
be invoked just like gcc, extract the appropriate macro information
based on options, and then invoke sparse.

I guess that's what cgcc is trying to be.  Since it also runs the compiler, no
special support for cgcc should be needed in Makefile other than using $(CC).

--
Regards,
Pavel Roskin