Richard Eckart wrote: > thanks for the backtrace. I heard there was a bug, but it never bit me > and I had no backtrace so far. > It should be fixed in current cvs now.
Sorry to interrupt, but isn't the bug rather calling hcache_remove_all() from hcache_free()? Also, why is host_count 24 if hostlist is NULL? hcache_remove_all() was already called for all cache types in hcache_close() in the previous loop. By explicitely checking whether the caches are NULL or not, you simply shadow the real bug. The code should also be changed to use a hash list. The manual list handling causes a unnecessary complexity and performance overhead. This should be directly possible by either using atoms for the hosts or modifying hashlist.c to support a custom comparison function - unless I missed something. -- Christian
pgpbRCgrtXyba.pgp
Description: PGP signature
