[Bug c++/58300] ICE: in decide_is_symbol_needed, at cgraphunit.c:233 with -fvtable-verify=preinit on invalid code

[cmtice at google dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58300

--- Comment #4 from Caroline Tice  ---
I am looking into this now.  I will update this bug when I know more.

[Bug c++/58300] ICE: in decide_is_symbol_needed, at cgraphunit.c:233 with -fvtable-verify=preinit on invalid code

[cmtice at google dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58300

--- Comment #6 from Caroline Tice  ---
When the preinit flag is used, the vtable verification constructor
initialization function was getting written to the assembly file before
cgraph_process_new_functions was being called (to process the new function),
which was causing an assertion failure in decide_is_symbol_needed (which
expects the symbol NOT to have its decl assembler name yet).  The fix is very
simple...reorder those two events.  This should not cause any errors.  I have a
patch that does this, which I have tested both on this test case and with the
new libvtv testsuite, and it passes in both places:

Index: gcc/cp/vtable-class-hierarchy.c
===
--- vtable-class-hierarchy.c(revision 202296)
+++ vtable-class-hierarchy.c(working copy)
@@ -1179,15 +1179,16 @@ vtv_generate_init_routine (void)
   TREE_USED (vtv_fndecl) = 1;
   DECL_PRESERVE_P (vtv_fndecl) = 1;
   if (flag_vtable_verify == VTV_PREINIT_PRIORITY)
-{
-  DECL_STATIC_CONSTRUCTOR (vtv_fndecl) = 0;
-  assemble_vtv_preinit_initializer (vtv_fndecl);
-}
+DECL_STATIC_CONSTRUCTOR (vtv_fndecl) = 0;

   gimplify_function_tree (vtv_fndecl);
   cgraph_add_new_function (vtv_fndecl, false);

   cgraph_process_new_functions ();
+
+  if (flag_vtable_verify == VTV_PREINIT_PRIORITY)
+assemble_vtv_preinit_initializer (vtv_fndecl);
+
 }
   pop_lang_context ();
 }


I will submit this patch for review soon.

[Bug c++/58300] ICE: in decide_is_symbol_needed, at cgraphunit.c:233 with -fvtable-verify=preinit on invalid code

[cmtice at google dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58300

--- Comment #7 from Caroline Tice  ---
Created attachment 30752
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=30752&action=edit
Reorder two function calls to prevent an ICE.

[Bug c++/58300] ICE: in decide_is_symbol_needed, at cgraphunit.c:233 with -fvtable-verify=preinit on invalid code

[cmtice at google dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58300

--- Comment #8 from Caroline Tice  ---
I have added the patch as an attachment, and also submitted it to the
gcc-patches list for review.

[Bug c++/58300] ICE: in decide_is_symbol_needed, at cgraphunit.c:233 with -fvtable-verify=preinit on invalid code

[cmtice at google dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58300

Caroline Tice  changed:

   What|Removed |Added

 Status|ASSIGNED|RESOLVED
 Resolution|--- |FIXED

--- Comment #10 from Caroline Tice  ---
See other comments.

[Bug other/58374] Wrong target check in configure.ac in libvtv

[cmtice at google dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58374

--- Comment #2 from Caroline Tice  ---
Even though I logged in to Bugzilla, I can't seem to edit any of the fields
above, but to the best of my knowledge a patch to fix this problem has been
committed to GCC and this bug should be marked as fixed and closed.

[Bug bootstrap/58441] [4.9 Regression] VTV headers are installed into the general include directory

[cmtice at google dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58441

--- Comment #3 from Caroline Tice  ---
Could you be a little more specific please?  What header are you seeing where?

[Bug bootstrap/65910] r222473 breaks x86_64 darwin bootstrap

[cmtice at google dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65910

--- Comment #4 from Caroline Tice  ---
Has anyone actually committed this fix?  I'm not seeing it in my tree yet

[Bug bootstrap/66521] xgcc: cc1plus segfaults when compiling libstdc++-v3/src/c++11/ostream-inst.cc

[cmtice at google dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66521

--- Comment #13 from Caroline Tice  ---
I'm looking at your patch now; assuming I don't find any problems with it, I
will try to get it committed.