Message ID | 20120425113818.GI26622@kam.mff.cuni.cz |
---|---|
State | New |
Headers | show |
Hi, the fortran problem is caused by fact that fortran does nested funtions that are static constructors. I did not really think of that case in cgraph construction code. Fixed thus. Honza PR middle-end/53089 * cgraphunit.c (cgraph_process_new_functions): Do not enqueue new functions here. (referred_to_p): Move ahead in file to avoid forward declaration. (cgraph_finalize_function): Finalize them here. * symtab.c (dump_symtab): Dump ctors and dtors. Index: cgraphunit.c =================================================================== *** cgraphunit.c (revision 186770) --- cgraphunit.c (working copy) *************** cgraph_process_new_functions (void) *** 243,249 **** cgraph_finalize_function (fndecl, false); output = true; cgraph_call_function_insertion_hooks (node); - enqueue_node ((symtab_node) node); break; case CGRAPH_STATE_IPA: --- 243,248 ---- *************** cgraph_reset_node (struct cgraph_node *n *** 320,325 **** --- 319,340 ---- cgraph_node_remove_callees (node); } + /* Return true when there are references to NODE. */ + + static bool + referred_to_p (symtab_node node) + { + int i; + struct ipa_ref *ref; + + for (i = 0; ipa_ref_list_referring_iterate (&node->symbol.ref_list, i, ref); + i++) + return true; + if (symtab_function_p (node) && cgraph (node)->callers) + return true; + return false; + } + /* DECL has been parsed. Take it, queue it, compile it at the whim of the logic in effect. If NESTED is true, then our caller cannot stand to have the garbage collector run at the moment. We would need to either create *************** cgraph_finalize_function (tree decl, boo *** 372,377 **** --- 387,397 ---- if (!nested) ggc_collect (); + + if (cgraph_state == CGRAPH_STATE_CONSTRUCTION + && (cgraph_decide_is_function_needed (node, decl) + || referred_to_p ((symtab_node)node))) + enqueue_node ((symtab_node)node); } /* Add the function FNDECL to the call graph. *************** process_function_and_variable_attributes *** 1114,1135 **** } } - /* Return true when there are references to NODE. */ - - static bool - referred_to_p (symtab_node node) - { - int i; - struct ipa_ref *ref; - - for (i = 0; ipa_ref_list_referring_iterate (&node->symbol.ref_list, i, ref); - i++) - return true; - if (symtab_function_p (node) && cgraph (node)->callers) - return true; - return false; - } - /* Mark DECL as finalized. By finalizing the declaration, frontend instruct the middle end to output the variable to asm file, if needed or externally visible. */ --- 1134,1139 ---- Index: symtab.c =================================================================== *** symtab.c (revision 186770) --- symtab.c (working copy) *************** dump_symtab_base (FILE *f, symtab_node n *** 414,419 **** --- 414,426 ---- fprintf (f, " virtual"); if (DECL_ARTIFICIAL (node->symbol.decl)) fprintf (f, " artificial"); + if (TREE_CODE (node->symbol.decl) == FUNCTION_DECL) + { + if (DECL_STATIC_CONSTRUCTOR (node->symbol.decl)) + fprintf (f, " constructor"); + if (DECL_STATIC_DESTRUCTOR (node->symbol.decl)) + fprintf (f, " destructor"); + } fprintf (f, "\n"); if (node->symbol.same_comdat_group)
On Wed, Apr 25, 2012 at 5:04 PM, Jan Hubicka <hubicka@ucw.cz> wrote: > + /* Return true when there are references to NODE. */ > + > + static bool > + referred_to_p (symtab_node node) > + { > + int i; > + struct ipa_ref *ref; > + > + for (i = 0; ipa_ref_list_referring_iterate (&node->symbol.ref_list, i, ref); > + i++) > + return true; This looks odd. Don't you want to do something with ref? > + if (symtab_function_p (node) && cgraph (node)->callers) > + return true; > + return false; > + } Ciao! Steven
> On Wed, Apr 25, 2012 at 5:04 PM, Jan Hubicka <hubicka@ucw.cz> wrote: > > + /* Return true when there are references to NODE. */ > > + > > + static bool > > + referred_to_p (symtab_node node) > > + { > > + int i; > > + struct ipa_ref *ref; > > + > > + for (i = 0; ipa_ref_list_referring_iterate (&node->symbol.ref_list, i, ref); > > + i++) > > + return true; > > This looks odd. Don't you want to do something with ref? No, I have just iteration API, not API to check how many refs are there and I want to see if there are any... Honza
> > On Wed, Apr 25, 2012 at 5:04 PM, Jan Hubicka <hubicka@ucw.cz> wrote: > > > + /* Return true when there are references to NODE. */ > > > + > > > + static bool > > > + referred_to_p (symtab_node node) > > > + { > > > + int i; > > > + struct ipa_ref *ref; > > > + > > > + for (i = 0; ipa_ref_list_referring_iterate (&node->symbol.ref_list, i, ref); > > > + i++) > > > + return true; > > > > This looks odd. Don't you want to do something with ref? > No, I have just iteration API, not API to check how many refs are there > and I want to see if there are any... Actually it is bit too much of cut&paste. Just if (ipa_ref_list_referring_iterate (&node->symbol.ref_list, 0, ref)) would work too. i will update it ;) Honza > > Honza
On 04/25/12 11:46, Jan Hubicka wrote: >>> On Wed, Apr 25, 2012 at 5:04 PM, Jan Hubicka <hubicka@ucw.cz> wrote: >>>> + /* Return true when there are references to NODE. */ >>>> + >>>> + static bool >>>> + referred_to_p (symtab_node node) >>>> + { >>>> + int i; >>>> + struct ipa_ref *ref; >>>> + >>>> + for (i = 0; ipa_ref_list_referring_iterate (&node->symbol.ref_list, i, ref); >>>> + i++) >>>> + return true; >>> >>> This looks odd. Don't you want to do something with ref? >> No, I have just iteration API, not API to check how many refs are there >> and I want to see if there are any... > Actually it is bit too much of cut&paste. Just > if (ipa_ref_list_referring_iterate (&node->symbol.ref_list, 0, ref)) > would work too. i will update it ;) And with a comment too, please. r~
Index: gcc.target/i386/pr39082-1.c =================================================================== --- gcc.target/i386/pr39082-1.c (revision 186814) +++ gcc.target/i386/pr39082-1.c (working copy) @@ -13,7 +13,7 @@ extern int bar1 (union un); extern union un bar2 (int); int -foo1 (union un u) +foo1 (union un u) /* { dg-message "note: the ABI of passing union with long double has changed in GCC 4.4" } */ { bar1 (u); return u.i; @@ -30,6 +30,6 @@ foo2 (void) int foo3 (int x) { - union un u = bar2 (x); /* { dg-message "note: the ABI of passing union with long double has changed in GCC 4.4" } */ + union un u = bar2 (x); return u.i; } Index: ChangeLog =================================================================== --- ChangeLog (revision 186814) +++ ChangeLog (working copy) @@ -1,3 +1,8 @@ +2012-04-25 Jan Hubicka <jh@suse.cz> + + PR middle-end/53088 + * gcc.target/i386/pr39082-1.c: Update warning location. + 2012-04-25 Jakub Jelinek <jakub@redhat.com> PR c/52880