On Fri, Nov 13, 2015 at 12:27:03PM +0100, Jakub Hrozek wrote:
> I have a different question. (i haven't read patches yet)
> But I can see that check_leaks_push is called after sysdb_store_user.
>
> I would like to know why.
> because we shout try to check leaks "caused" in this function.
Wouldn't these leaks be caught by leaks checks that are pushed in
setup() and popped in teardown() ?
Ah, but we don't use leak checks in setup() and teardown(), sorry!
Then it's something we definitely need to fix -- I think a leak in
requests (like cached_req) is only possible if we use a long-lived
(responder or event) context by mistake.
So I think we should file another ticket and add leak checks to setup and
teardown in this setup. It's not related to Petr's tests, but I think it
should be done before we switch to the cache_req API in NSS to make sure
we don't regress.