| Message ID | 20100628134736.GC4383@virgil.arch.suse.de |
|---|---|
| State | New |
| Headers | show |
> Hi, > > when I was going through the IPA-CP analysis code again I noticed that > ipcp_init_cloned_node was calling some initialization functions that > have already been done by node duplication hook (and the > initialization functions then do nothing because they check whether > the data structures they allocate already exist). The following patch > therefore replaces them with an assert. I decided that checking > assert would do because if the condition is false, we would most most > probably run into a segfaut later. > > Bootstrapped and tested on x86_64-linux without any issues. OK for > trunk? > > Thanks, > > Martin > > 2010-06-27 Martin Jambor <mjambor@suse.cz> > > * ipa-cp.c (ipcp_init_cloned_node): Replace calls to > ipa_check_create_node_params and ipa_initialize_node_params with > checking asserts they are not necessary. OK. Honza
Index: icln/gcc/ipa-cp.c =================================================================== --- icln.orig/gcc/ipa-cp.c +++ icln/gcc/ipa-cp.c @@ -172,8 +172,11 @@ static void ipcp_init_cloned_node (struct cgraph_node *orig_node, struct cgraph_node *new_node) { - ipa_check_create_node_params (); - ipa_initialize_node_params (new_node); + gcc_checking_assert (ipa_node_params_vector + && (VEC_length (ipa_node_params_t, + ipa_node_params_vector) + > (unsigned) cgraph_max_uid)); + gcc_checking_assert (IPA_NODE_REF (new_node)->params); IPA_NODE_REF (new_node)->ipcp_orig_node = orig_node; }