New and bogus(!) warning produced by sparse.

New and bogus(!) warning produced by sparse.

[Anton Altaparmakov]

Hi Linus,

I just did a git pull on sparse and it produced a new and I believe 
bogus warning when trying to assign a value to a bit field member.  The 
warning is:

warning: generating address of non-lvalue (1)

Below is an example program that triggers the warning when you run:

	sparse sparse-test.c

Best regards,

	Anton
-- 
Anton Altaparmakov <aia21 at cam.ac.uk> (replace at with @)
Unix Support, Computing Service, University of Cambridge, CB2 3QH, UK
Linux NTFS maintainer / IRC: #ntfs on irc.freenode.net
WWW: http://linux-ntfs.sf.net/ & http://www-stu.christs.cam.ac.uk/~aia21/

--- sparse-test.c ---

int main(void)
{
	struct {
		unsigned bit1:1;
		unsigned bit2:1;
	} bits = { 0, 0 };

	bits.bit1 = 1; /* sparse warns here! */
	bits.bit2 = 0; /* sparse warns here! */
	return 0;
}

Re: New and bogus(!) warning produced by sparse.

[Linus Torvalds]

On Fri, 30 Sep 2005, Anton Altaparmakov wrote:
> 
> I just did a git pull on sparse and it produced a new and I believe 
> bogus warning when trying to assign a value to a bit field member.  The 
> warning is:

Thanks, that's indeed very bogus. 

I see why it happens, will fix sparse momentarily (btw, for sparse bugs, 
use "linux-sparse@vger.kernel.org" rather than LKML - since this is 
clearly not a kernel problem ;)

		Linus

Re: New and bogus(!) warning produced by sparse.

[Linus Torvalds]

On Fri, 30 Sep 2005, Linus Torvalds wrote:
> 
> I see why it happens, will fix sparse momentarily (btw, for sparse bugs, 
> use "linux-sparse@vger.kernel.org" rather than LKML - since this is 
> clearly not a kernel problem ;)

Side note: it has nothing to do with bit-fields: it happens for any 
structure member that sparse will have incorrectly figured out as having a 
value pre-assigned to it. So even

	struct {
		int value;
	} dummy = { 0 };

	dummy.value = 1;

will generate the warning. It isn't that a bitfield cannot have it's 
address taken, it's that sparse has incorrectly "optimized" dummy.value to 
be 0, so it ends up evaluating it as

	0 = 1;

and says that it's trying to assign a non-lvalue.. Very stupid.

		Linus

Re: New and bogus(!) warning produced by sparse.

[Linus Torvalds]

On Fri, 30 Sep 2005, Linus Torvalds wrote:
> 
> I see why it happens, will fix sparse momentarily 

Fixed, pushed out.

		Linus

Re: New and bogus(!) warning produced by sparse.

[Anton Altaparmakov]

On Fri, 2005-09-30 at 08:37 -0700, Linus Torvalds wrote:
> 
> On Fri, 30 Sep 2005, Linus Torvalds wrote:
> > 
> > I see why it happens, will fix sparse momentarily 
> 
> Fixed, pushed out.

Pulled, compiled and warning is gone.  Thanks!

And sorry I sent it to lkml, I had forgotten about linux-sparse ml and
it started out as a kernel related email given I saw it when compiling
kernel source.  (-;

Best regards,

        Anton
-- 
Anton Altaparmakov <aia21 at cam.ac.uk> (replace at with @)
Unix Support, Computing Service, University of Cambridge, CB2 3QH, UK
Linux NTFS maintainer / IRC: #ntfs on irc.freenode.net
WWW: http://linux-ntfs.sf.net/ & http://www-stu.christs.cam.ac.uk/~aia21/