This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: PR c++/25010


Mark Mitchell <mark@codesourcery.com> writes:

| Andrew Pinski wrote:
| >>In general, the optimizers need to be careful to interpret
| >>DECL_EXTERNAL as meaning "do not output in this translation unit", and
| >>not as "no definition is available in this translation unit".
| > 
| > 
| > tree.h has this comment:
| > /* In a VAR_DECL or FUNCTION_DECL, nonzero means external reference:
| >    do not allocate storage, and refer to a definition elsewhere.  */
| > 
| > which means no definition is avilable in this TU or at least the way
| > I am reading it.
| 
| Maybe the documentation should change, but the way you are reading it is
| certainly not what DECL_EXTERNAL means, or has meant in some time.

I, too, am confused about what DECL_EXTERNAL really means.  With
-fdump-translation-unit, I regularly see

     body: undefined

because DECL_EXTERNAL was set, see tree-dump.c:

    case FUNCTION_DECL:
      dump_child ("args", DECL_ARGUMENTS (t));
      if (DECL_EXTERNAL (t))
        dump_string_field (di, "body", "undefined");

and later, 

    body: @xxx

because DECL_SAVED_TREE is also set

      if (DECL_LANG_SPECIFIC (t) && !dump_flag (di, TDF_SLIM, t))
        dump_child ("body", DECL_SAVED_TREE (t));

So, there seems to be an inconsistent practice in the compiler; I
don't which is right and which is wrong.

-- Gaby


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]