nit in hcb_deref?

[Nils Goroll]

Trying to understand if I race a saw in 2.0.x still exists in master, I reviewed
some of the cash_hash.c and hash_*.c code.

The issue I feared could exist seems not to, but I believe I spottet a little nit:

* HSH_DeleteObjHead is only called if hash_deref(oh) returned 0

  - hcb_deref always returns 1, so for hcb we never call

	Lck_Delete(&oh->mtx);
	ds->n_objecthead--;
	FREE_OBJ(oh);

    looks like a resource leak to me

    simple fix IMHO:


diff --git a/bin/varnishd/hash/hash_critbit.c b/bin/varnishd/hash/hash_critbit.c
index 4e14bdd..d27e3ce 100644
--- a/bin/varnishd/hash/hash_critbit.c
+++ b/bin/varnishd/hash/hash_critbit.c
@@ -408,6 +408,7 @@ hcb_deref(struct objhead *oh)
                Lck_Unlock(&hcb_mtx);
                assert(VTAILQ_EMPTY(&oh->objcs));
                AZ(oh->waitinglist);
+               r = 0;
        }
        Lck_Unlock(&oh->mtx);
 #ifdef PHK


  - hsl_deref / hcl_deref look ok to me

Nils