[Bug preprocessor/33305] New: We should warn about empty macro arguments

[Bug preprocessor/33305] New: We should warn about empty macro arguments

[ghazi at gcc dot gnu dot org]

As noted in PR33304, empty macro arguments can cause problems for ISO C90
compilers (solaris cc) as well as for older gcc versions like 2.8.1.  Therefore
I'd like to see cpp warn about these constructs either with -pedantic in C90
more and/or its own flag.

We should use this warning during bootstrap to keep the sources bootstrappable
when using compilers that have this issue.  We have to fix the existing cases
before we do that though.


-- 
           Summary: We should warn about empty macro arguments
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Severity: enhancement
          Priority: P3
         Component: preprocessor
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: ghazi at gcc dot gnu dot org
 BugsThisDependsOn: 33304


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33305

[Bug preprocessor/33305] We should warn about empty macro arguments

[manu at gcc dot gnu dot org]

------- Comment #1 from manu at gcc dot gnu dot org  2007-09-05 09:24 -------
There was talking about creating a -Wundefined flag that warns about undefined
behaviour (PR30334). Would this fit in there? (-pedantic is not supposed to
warn about undefined constructions as far as I know).


-- 

manu at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |manu at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33305

[Bug preprocessor/33305] We should warn about empty macro arguments

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2007-09-05 09:29 -------
Subject: Re:  We should warn about empty macro arguments

On 5 Sep 2007 09:24:09 -0000, manu at gcc dot gnu dot org
<gcc-bugzilla@gcc.gnu.org> wrote:
> There was talking about creating a -Wundefined flag that warns about undefined
> behaviour (PR30334). Would this fit in there? (-pedantic is not supposed to
> warn about undefined constructions as far as I know).

There are two different undefined behaviors, one at runtime and one at
compile time.  This belongs to the latter which means we can
diagnostic this if we want (and can do it with -pedantic in fact).

-- Pinski


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33305

Re: [Bug preprocessor/33305] We should warn about empty macro arguments

[Andrew Pinski]

On 5 Sep 2007 09:24:09 -0000, manu at gcc dot gnu dot org
<gcc-bugzilla@gcc.gnu.org> wrote:
> There was talking about creating a -Wundefined flag that warns about undefined
> behaviour (PR30334). Would this fit in there? (-pedantic is not supposed to
> warn about undefined constructions as far as I know).

There are two different undefined behaviors, one at runtime and one at
compile time.  This belongs to the latter which means we can
diagnostic this if we want (and can do it with -pedantic in fact).

-- Pinski

[Bug preprocessor/33305] We should warn about empty macro arguments

[ghazi at gcc dot gnu dot org]

------- Comment #3 from ghazi at gcc dot gnu dot org  2007-09-05 17:29 -------
Anyone know where in the preprocessor this should be done?  I.e. which function
in libcpp?

If someone would let me know, I'll write a patch.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33305

[Bug preprocessor/33305] We should warn about empty macro arguments

[manu at gcc dot gnu dot org]

------- Comment #4 from manu at gcc dot gnu dot org  2007-11-09 02:09 -------
(In reply to comment #3)
> Anyone know where in the preprocessor this should be done?  I.e. which function
> in libcpp?
> 
> If someone would let me know, I'll write a patch.
> 

This is the patch I was playing with before giving up. I don't know how to get
the correct line numbers so I tried 3 possibilities.

Index: libcpp/macro.c
===================================================================
--- libcpp/macro.c      (revision 129897)
+++ libcpp/macro.c      (working copy)
@@ -681,9 +681,39 @@ collect_args (cpp_reader *pfile, const c
     }
   else
     {
+      unsigned int i;
+
       /* A single empty argument is counted as no argument.  */
       if (argc == 1 && macro->paramc == 0 && args[0].count == 0)
        argc = 0;
+
+      for (i = 0; i < argc; i++)
+        {
+          if (args[i].count == 0
+              && CPP_OPTION (pfile, pedantic)
+              && CPP_OPTION (pfile, lang) == CLK_GNUC89)
+            {
+              if (pfile->context->prev && pfile->context->prev->macro)
+                cpp_error_with_line (pfile, CPP_DL_PEDWARN,
+                    pfile->context->prev->macro->value.macro->line, 0,
+                    "invoking  macro \"%s\" "
+                    "with empty arguments is undefined in ISO C90",
+                    NODE_NAME (node));
+              else if (pfile->context->prev || pfile->state.in_directive)
+                cpp_error_with_line (pfile, CPP_DL_PEDWARN,
+                                     node->value.macro->line, 0,
+                                     "invoking  macro \"%s\" "
+                    "with empty arguments is undefined in ISO C90",
+                    NODE_NAME (node));
+              else
+                cpp_error (pfile, CPP_DL_PEDWARN,
+                           "invoking  macro \"%s\" "
+                           "with empty arguments is undefined in ISO C90",
+                           NODE_NAME (node));
+
+          }
+        }
+
       if (_cpp_arguments_ok (pfile, macro, node, argc))
        {
          /* GCC has special semantics for , ## b where b is a varargs

And this is the testcase I am using:

/* { dg-do compile } */
/* { dg-options "-std=c90 -pedantic-errors" } */

#define f2(a,b,c) a; b; c;
#define f(a,b)   f2(a,,b) /* { dg-warning "macro "f2" with empty arguments } */
#define f3(a)    a

#define g()   0

void foo(void)
{
    f(0,0);
    f2(0,,0); /* { dg-warning "macro "f2" with empty arguments } */
    f3(); /* { dg-warning "macro "f3" with empty arguments } */
    g();
}


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33305

[Bug preprocessor/33305] We should warn about empty macro arguments

[aph at gcc dot gnu dot org]

------- Comment #5 from aph at gcc dot gnu dot org  2008-07-03 10:14 -------
Subject: Bug 33305

Author: aph
Date: Thu Jul  3 10:13:48 2008
New Revision: 137411

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=137411
Log:
2008-07-03  Andrew Haley  <aph@redhat.com>

        PR preprocessor/33305
        * gcc.dg/cpp/avoidpaste1.c: Use dg-options "-ansi" to avoid
        "-pedantic".
        * gcc.dg/cpp/avoidpaste2.c: Likewise
        * gcc.dg/cpp/20000519-1.c: Likewise.
        * g++.dg/ext/gnu-inline-global-reject.C: Likewise.
        * gcc.dg/cpp/c99-empty-macro-args.c: New test.
        * gcc.dg/cpp/c90-empty-macro-args.c: New test.


Added:
    trunk/gcc/testsuite/gcc.dg/cpp/c90-empty-macro-args.c
    trunk/gcc/testsuite/gcc.dg/cpp/c99-empty-macro-args.c
Modified:
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/testsuite/g++.dg/ext/gnu-inline-global-reject.C
    trunk/gcc/testsuite/gcc.dg/cpp/20000519-1.c
    trunk/gcc/testsuite/gcc.dg/cpp/avoidpaste1.c
    trunk/gcc/testsuite/gcc.dg/cpp/avoidpaste2.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33305

[Bug preprocessor/33305] We should warn about empty macro arguments

[aph at gcc dot gnu dot org]

------- Comment #6 from aph at gcc dot gnu dot org  2008-07-03 10:32 -------
Subject: Bug 33305

Author: aph
Date: Thu Jul  3 10:31:50 2008
New Revision: 137414

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=137414
Log:
2008-06-13  Andrew Haley  <aph@redhat.com>

        PR preprocessor/33305
        * macro.c (replace_args): Print a warning for empty macro
        arguments in C89 and C++.


Modified:
    trunk/libcpp/ChangeLog
    trunk/libcpp/macro.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33305

[Bug preprocessor/33305] We should warn about empty macro arguments

[tromey at gcc dot gnu dot org]

------- Comment #7 from tromey at gcc dot gnu dot org  2008-07-12 17:06 -------
Fixed on trunk.


-- 

tromey at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|---                         |4.4.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33305

[Bug preprocessor/33305] We should warn about empty macro arguments

[vz-gcc at zeitlins dot org]

------- Comment #8 from vz-gcc at zeitlins dot org  2010-03-17 20:05 -------
Sorry for a late follow up but I've just discovered that this change broke
compilation of code using wxWidgets library with "-pedantic-errors -std=c++98"
switches because wxWidgets uses constructions such as (simplified):

#define wxEMPTY_PARAMETER_VALUE /* Fake macro parameter value */

/*
This will be __declspec(dllexport) for some compilers, __attribute__
((visibility("default"))) for other ones and empty in non-shared library build
*/
#define wxEXPORT ...

#define wxDECLARE_SOMETHING(expdecl, name) class expdecl name
#define wxDECLARE_EXPORTED_SOMETHING(name) wxDECLARE_SOMETHING(wxEXPORT, name)


And now, with g++ 4.4.3 from Debian Unstable, this results in "empty macro
arguments are undefined in ISO C90 and ISO C++98" warning/error when wxEXPORT
is empty.

Admittedly, the error may be correct (although to be honest I am not even 100%
sure it is even after rereading 16.3 and 16.3.1 of C++98 several times), but
this use is common and often occurs in other code. Even Boost defines
BOOST_PP_EMPTY in its boost/preprocessor/facilities/empty.hpp and I definitely
saw this many times in other code as well so at the very least there is a
widespread belief in C++ community that such usage is valid.

So ideally I'd like to see gcc behave more leniently when a macro which
expands to an empty value is passed to another macro as argument (as opposed
to directly omitting a macro argument). Failing that, it'd be nice to have a
specific option to turn this behaviour off. Because right now the only choice
seems to be to either disable -pedantic (which would be a really unfortunate
consequence of the change making gcc stricter) or use -std=c++0x which is more
acceptable but still not always desirable.

TIA!


-- 

vz-gcc at zeitlins dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |vz-gcc at zeitlins dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33305