error: ctype on uninitialized symbol

error: ctype on uninitialized symbol

[Bob Copeland]

Hi all,

I'm seeing the following error when running sparse with latest HEAD,
against the small test case at the end:

$ sparse sparse-test.c
[garbage]:0:0: error: ctype on uninitialized symbol 0x696ec0

I am getting this with calls to time_after(..., msecs_to_jiffies()) in my
kernel build.

Below, the "1UL + 1L" is a stand-in for MAX_JIFFY_OFFSET, and the value
has to be some operation on two different types to trigger for me (e.g.
1UL + 1UL will run without error).  Taking out the typecheck macro will
also remove the error.

Any thoughts?

$ sparse --version
v0.5.0-44-g40791b9

sparse-test.c:

#define typecheck(type,x) \
({	type __dummy; \
	typeof(x) __dummy2; \
	(void)(&__dummy == &__dummy2); \
	1; \
})

#define time_after(a,b)		\
	(typecheck(unsigned long, a) && \
	 typecheck(unsigned long, b) && \
	 ((long)((b) - (a)) < 0))

static inline unsigned long y(void)
{
	return 1UL+1L;
}

static void x(void)
{
	int c = time_after(1UL, y());
}

-- 
Bob Copeland %% http://bobcopeland.com/

Re: error: ctype on uninitialized symbol

[Bob Copeland]

On Mon, Feb 01, 2016 at 02:44:09PM -0500, Bob Copeland wrote:
> Hi all,
> 
> I'm seeing the following error when running sparse with latest HEAD,
> against the small test case at the end:
> 
> $ sparse sparse-test.c
> [garbage]:0:0: error: ctype on uninitialized symbol 0x696ec0

Nevermind, I fooled myself.  Turns out this "sparse" was an old version
of smatch, not the freshly-built sparse that "sparse --version" was
reporting.

It looks like this bug was fixed around 0.4.4.

Sorry for the noise!

-- 
Bob Copeland %% http://bobcopeland.com/