> Thanks for the bug report. I was just working on that code last week
> and I just pushed a fix right now.
>
> commit 105d6766f7f2 ('implied: remove buggy code for working around the old fake_stree methods')
This fixed problems mentioned in my first mail but introduces new ones:
cat << EOF > test.c
struct bar {
void *x;
};
void doit(void);
int checkit(void);
struct bar *getit(void);
#define WARNO(condition, format...) ({ \
int __ret_warn_on = !!(condition); \
if (__ret_warn_on) \
doit(); \
__ret_warn_on; \
})
struct lhead {
struct lhead *next, *prev;
};
#define lfor_each(pos, head) \
for (pos = (head)->next; pos != (head); pos = pos->next)
void foo(struct lhead *head)
{
struct lhead *h;
struct lhead *h2 = (void *)0;
struct bar *a;
struct bar *b = (void *)0;
lfor_each(h, head) {
if (checkit()) {
if (WARNO(h2, "foobaro"))
doit();
h2 = h;
}
}
a = getit();
if (a)
b = getit();
if (b)
b->x = a->x;
}
EOF
smatch -p=kernel --two-passes test.c
This minimized version looks quite odd but it is even more odd that removing
a single unrelated looking line (like for example the lfor_each stuff or the
WARNO) will "fix" the warning:
test.c:45 foo() error: we previously assumed 'a' could be null (see line 41)
--
To unsubscribe from this list: send the line "unsubscribe smatch" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
