On 22/07/15 08:58, Thomas Gleixner wrote:

On Wed, 22 Jul 2015, Marc Zyngier wrote:

quoted

Looks like my coccinelle foo is still lacking a bit... Oh well.

Please post your semantic patch and don't forget to Cc Julia Lawall :)

Bah, found the problem. My semantic patch is extremely simple:

@init@
struct irq_domain *ptr;
position pos;
@@

(
ptr at pos->of_node
)


@ script:python @
p1 << init.pos;
@@

for p in p1:
    print "%s:%s" % (p.file,p.line)

as I'm just grepping through the tree (I don't feel confident enough to
do the patching, and some cases are a bit more complicated).

But I forgot to add arch/$ARCH/include/asm as an include search path for
spatch. That had the side effect of spatch being unable to identify
constructs like:

arch/$ARCH/include/asm/foo.h:

	struct foo {
		struct irq_domain *bar;
	};

arch/$ARCH/.../foo.c:

	struct foo foo;
	if (foo->bar->of_node) { ... }

Since spatch knows nothing about struct foo, it cannot identify bar as a
struct irq_domain, hence missing this occurrence.

Once I fixed the include path, I found an additional couple of victims.
I really love this thing! :-)

	M.
-- 
Jazz is not dead. It just smells funny...