Re: [PATCH] Fix ICE with -Wodr (PR middle-end/61913)

[Jan Hubicka <hubicka at ucw dot cz>]

> On Mon, Jul 28, 2014 at 10:23:36AM +0200, Jan Hubicka wrote:
> > > On Sun, Jul 27, 2014 at 1:02 PM, Marek Polacek <polacek@redhat.com> wrote:
> > > > Wodr in common.opt was missing a Var, which means:
> > > > 1) we ICE with -Wodr, since -Wodr isn't handled in opts.c;
> > > > 2) -Wno-odr wouldn't work.
> > > > Thus fixed.  I'd think this doesn't need a testcase...
> > > >
> > > > Bootstrapped/regtested on x86_64-linux, ok for trunk?
> > > 
> > > Ok.  Does this mean we don't have a -Wodr testcase?
> > 
> > We ICE at -Wno-odr (thanks for fixing it! I was somewhat puzzled by fact
> > that I need to declare var that is not needed at all).
> > 
> > But there is no -Wodr testcase as I do not know how to match warnings at LTO
> > time. If there is a way, I will add one.
> 
> Heh, I was just trying to create some testcase for -Wodr, but failed.
> pr60720_0.c testcase says:
> 
> /* ???  lto.exp does not allow to scan for
>    :1:12: warning: type of 'x' does not match original declaration
>     extern int x[];
>                ^
>    :1:5: note: previously declared here
>     int x;
>         ^  */

This is not ODR warning however, it is declaration mismatch.  Yes, lack of
scanning machinery is sad ;( BTW these type of warnings tends ot be
uninformative in the sense:

>    :1:5: warning: type of 'x' does not match original declaration
>     foo x;
>                ^
>    :1:5: note: previously declared here
>     foo x;

That is caused by different defines or typedefs earlier in the file.  This is
uninformative enough so users tends to ignore the warning as obvious GCC bug.
I tested attached patch that adds unit name into TRANSLATION_UNIT_DECL. It adds
main source file, perhaps output object file name would make more sense, but I
did not find global var for that.

Then I use it as follows:
  /* See if we have info about the translation unit.  It may not be around
     if types was already merged.   */
  while (TREE_CODE (name) != TRANSLATION_UNIT_DECL)
    name = TYPE_P (name) ? TYPE_CONTEXT (name) : DECL_CONTEXT (name);
  while (TREE_CODE (name1) != TRANSLATION_UNIT_DECL)
    name1 = TYPE_P (name1) ? TYPE_CONTEXT (name1) : DECL_CONTEXT (name1);
  name = DECL_NAME (name);
  name1 = DECL_NAME (name1);
  if (name != name1 && name && name1)
    inform (UNKNOWN_LOCATION, "Conflicting compilation units: %s and %s",
            IDENTIFIER_POINTER (name),
            IDENTIFIER_POINTER (name1));

We may want to have some sort of convenience wrapper for this.

Because the type merging ignores TRANSLATION_UNIT_DECL, it may end up being
wrong.  In the contextes we use it however - that we have unmerged duplicated
decls or unmerged duplicated types, it however usually points to two different
files that are the origins of the trees that is good enough.

Index: tree.c
===================================================================
--- tree.c	(revision 212987)
+++ tree.c	(working copy)
@@ -4539,6 +4540,8 @@ build_translation_unit_decl (tree name)
   tree tu = build_decl (UNKNOWN_LOCATION, TRANSLATION_UNIT_DECL,
 			name, NULL_TREE);
   TRANSLATION_UNIT_LANGUAGE (tu) = lang_hooks.name;
+  if (main_input_filename)
+    DECL_NAME (tu) = get_identifier (main_input_filename);
   vec_safe_push (all_translation_units, tu);
   return tu;
 }