On Tue, Oct 22, 2019 at 01:06:28PM +0100, Ben Dooks wrote: > On 25/09/2019 11:00, Ben Dooks wrote: > > This is a re-send of the printf formatting parsing to see if there is > > anything else that needs doing before getting it added to the next > > release. > > I've put a new copy of this up at: > > https://github.com/bjdooks-ct/sparse bjdooks/printf19 Thank you for the quick updates. I think everything is covered but my 2 last emails: * an update about string-- in evaluate_format_printf(). a test case is the following: void prt(const char *, ...) __attribute__((format(printf, 1, 2))); static void foo(int x, int y, const void *p) { prt("%02x%02x %8p\n", x, y, p); } * one about get_printf_fmt() and the dereference of expr->string and expression's type vs. expr->symbol. A testcase for it is: void prt(const char *, ...) __attribute__((format(printf, 1, 2))); static inline void fun(int x) { prt("%08x\n", x); } static void foo(int x) { fun; fun(x); } If you could also add to your tests one like: void prt(const char *, ...) __attribute__((format(printf, 1, 2))); static void foo(int i, unsigned int u) { prt("%d\n", i); prt("%x\n", u); } This is surprisingly not covered by your more complex ones. > we're back to having some tests fail as it looks like the current > evaluation does not care about int vs long. Yes. Sorry if I wasn't clear enough but I said that it needs something *like* check_assignment_types() instead of directly comparing the pointers. But, yes, check_assignment_types() itself ignore the size and signedness. If help is needed, I can look at this in a few days. -- Luc
