[C++ diagnostic] Handle MEMBER_REF/DOTSTAR_EXPR in C++ dump_expr (PR c++/33844)

[C++ diagnostic] Handle MEMBER_REF/DOTSTAR_EXPR in C++ dump_expr (PR c++/33844)

[Jakub Jelinek]

Hi!

pp_cxx_expression already handles these two (though incorrectly for
MEMBER_REF), but dump_expr does not.
Ok for trunk?

2007-10-26  Jakub Jelinek  <jakub@redhat.com>

	PR c++/33844
	* cxx-pretty-print.c (pp_cxx_pm_expression) <case MEMBER_REF>: Print
	->* rather than .*.
	* error.c (dump_expr): Handle MEMBER_REF and DOTSTAR_EXPR.

	* g++.dg/other/ptrmem8.C: New test.

--- gcc/cp/cxx-pretty-print.c.jj	2007-10-26 17:59:48.000000000 +0200
+++ gcc/cp/cxx-pretty-print.c	2007-10-26 18:07:13.000000000 +0200
@@ -814,7 +814,10 @@ pp_cxx_pm_expression (cxx_pretty_printer
     case MEMBER_REF:
     case DOTSTAR_EXPR:
       pp_cxx_pm_expression (pp, TREE_OPERAND (t, 0));
-      pp_cxx_dot (pp);
+      if (TREE_CODE (t) == MEMBER_REF)
+	pp_cxx_arrow (pp);
+      else
+	pp_cxx_dot (pp);
       pp_star(pp);
       pp_cxx_cast_expression (pp, TREE_OPERAND (t, 1));
       break;
--- gcc/cp/error.c.jj	2007-10-26 16:20:28.000000000 +0200
+++ gcc/cp/error.c	2007-10-26 18:06:15.000000000 +0200
@@ -2052,6 +2052,11 @@ dump_expr (tree t, int flags)
       pp_cxx_offsetof_expression (cxx_pp, t);
       break;
 
+    case MEMBER_REF:
+    case DOTSTAR_EXPR:
+      pp_multiplicative_expression (cxx_pp, t);
+      break;
+
     case DELETE_EXPR:
     case VEC_DELETE_EXPR:
       pp_cxx_delete_expression (cxx_pp, t);
--- gcc/testsuite/g++.dg/other/ptrmem8.C.jj	2007-10-26 18:08:53.000000000 +0200
+++ gcc/testsuite/g++.dg/other/ptrmem8.C	2007-10-26 18:14:30.000000000 +0200
@@ -0,0 +1,16 @@
+// PR c++/33844
+// { dg-do compile }
+
+struct A {};
+
+template<int> void foo(void (A::* f)())
+{
+  A a;
+  &(a.*f);	// { dg-error "invalid use of\[^\n\]*\\.\\*\[^\n\]*to form|qualified-id is required" }
+}
+
+template<int> void bar(void (A::* f)())
+{
+  A *p;
+  &(p->*f);	// { dg-error "invalid use of\[^\n\]*->\\*\[^\n\]*to form|qualified-id is required" }
+}

	Jakub

Re: [C++ diagnostic] Handle MEMBER_REF/DOTSTAR_EXPR in C++ dump_expr (PR c++/33844)

[Paolo Carlini]

Hi,
> --- gcc/testsuite/g++.dg/other/ptrmem8.C.jj	2007-10-26 18:08:53.000000000 +0200
> +++ gcc/testsuite/g++.dg/other/ptrmem8.C	2007-10-26 18:14:30.000000000 +0200
> @@ -0,0 +1,16 @@
> +// PR c++/33844
> +// { dg-do compile }
> +
> +struct A {};
> +
> +template<int> void foo(void (A::* f)())
> +{
> +  A a;
> +  &(a.*f);	// { dg-error "invalid use of\[^\n\]*\\.\\*\[^\n\]*to form|qualified-id is required" }
> +}
> +
> +template<int> void bar(void (A::* f)())
> +{
> +  A *p;
> +  &(p->*f);	// { dg-error "invalid use of\[^\n\]*->\\*\[^\n\]*to form|qualified-id is required" }
> +}
>   
In the output I'm seeing:

‘foo::a.*foo::f'

and

‘foo::p->*foo::f’

are we really sure we want to qualify with foo:: like that?

Thanks,
Paolo.

Re: [C++ diagnostic] Handle MEMBER_REF/DOTSTAR_EXPR in C++ dump_expr (PR c++/33844)

[Gabriel Dos Reis]

Paolo Carlini <pcarlini@suse.de> writes:

| Hi,
| > --- gcc/testsuite/g++.dg/other/ptrmem8.C.jj	2007-10-26 18:08:53.000000000 +0200
| > +++ gcc/testsuite/g++.dg/other/ptrmem8.C	2007-10-26 18:14:30.000000000 +0200
| > @@ -0,0 +1,16 @@
| > +// PR c++/33844
| > +// { dg-do compile }
| > +
| > +struct A {};
| > +
| > +template<int> void foo(void (A::* f)())
| > +{
| > +  A a;
| > +  &(a.*f);	// { dg-error "invalid use of\[^\n\]*\\.\\*\[^\n\]*to form|qualified-id is required" }
| > +}
| > +
| > +template<int> void bar(void (A::* f)())
| > +{
| > +  A *p;
| > +  &(p->*f);	// { dg-error "invalid use of\[^\n\]*->\\*\[^\n\]*to form|qualified-id is required" }
| > +}
| >   
| In the output I'm seeing:
| 
| ‘foo::a.*foo::f'
| 
| and
| 
| ‘foo::p->*foo::f’
| 
| are we really sure we want to qualify with foo:: like that?

No, one would need to set a `scope' cutoff.

-- Gaby

Re: [C++ diagnostic] Handle MEMBER_REF/DOTSTAR_EXPR in C++ dump_expr (PR c++/33844)

[Jakub Jelinek]

On Fri, Oct 26, 2007 at 01:34:02PM -0500, Gabriel Dos Reis wrote:
> No, one would need to set a `scope' cutoff.

So should reinit_cxx_pp set enclosing_scope to current_function_decl
instead of 0?

	Jakub

Re: [C++ diagnostic] Handle MEMBER_REF/DOTSTAR_EXPR in C++ dump_expr (PR c++/33844)

[Gabriel Dos Reis]

On Fri, 26 Oct 2007, Jakub Jelinek wrote:

| On Fri, Oct 26, 2007 at 01:34:02PM -0500, Gabriel Dos Reis wrote:
| > No, one would need to set a `scope' cutoff.
| 
| So should reinit_cxx_pp set enclosing_scope to current_function_decl
| instead of 0?

Yes, try it out if it works.

-- Gaby

Re: [C++ diagnostic] Handle MEMBER_REF/DOTSTAR_EXPR in C++ dump_expr (PR c++/33844)

[Jakub Jelinek]

On Fri, Oct 26, 2007 at 03:05:09PM -0500, Gabriel Dos Reis wrote:
> On Fri, 26 Oct 2007, Jakub Jelinek wrote:
> 
> | On Fri, Oct 26, 2007 at 01:34:02PM -0500, Gabriel Dos Reis wrote:
> | > No, one would need to set a `scope' cutoff.
> | 
> | So should reinit_cxx_pp set enclosing_scope to current_function_decl
> | instead of 0?
> 
> Yes, try it out if it works.

With the additional patch below other/ptrmem8.C reports:
/usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C: In function 'void foo(void (A::*)())':
/usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C:9: error: invalid use of 'a.*f' to form a pointer-to-member-function
/usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C:9: note:   a qualified-id is required
/usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C: In function 'void bar(void (A::*)())':
/usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C:15: error: invalid use of 'p->*f' to form a pointer-to-member-function
/usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C:15: note:   a qualified-id is required
and no regressions appear in regression testing.
Ok to commit both patches to the trunk?

2007-10-27  Jakub Jelinek  <jakub@redhat.com>

	* error.c (reinit_cxx_pp): Initialize cxx_pp->enclosing_scope
	to current_function_decl rather than 0.

--- gcc/cp/error.c.jj	2007-10-26 18:06:15.000000000 +0200
+++ gcc/cp/error.c	2007-10-27 11:52:10.000000000 +0200
@@ -2107,7 +2107,7 @@ reinit_cxx_pp (void)
   pp_base (cxx_pp)->padding = pp_none;
   pp_indentation (cxx_pp) = 0;
   pp_needs_newline (cxx_pp) = false;
-  cxx_pp->enclosing_scope = 0;
+  cxx_pp->enclosing_scope = current_function_decl;
 }
 
 
	Jakub

Re: [C++ diagnostic] Handle MEMBER_REF/DOTSTAR_EXPR in C++ dump_expr (PR c++/33844)

[Paolo Carlini]

> 2007-10-27  Jakub Jelinek  <jakub@redhat.com>
>
> 	* error.c (reinit_cxx_pp): Initialize cxx_pp->enclosing_scope
> 	to current_function_decl rather than 0.
>   
Thanks Jakub for this, really an unrelated issue, but time to fix it!

Paolo.

Re: [C++ diagnostic] Handle MEMBER_REF/DOTSTAR_EXPR in C++ dump_expr (PR c++/33844)

[Gabriel Dos Reis]

Jakub Jelinek <jakub@redhat.com> writes:

| On Fri, Oct 26, 2007 at 03:05:09PM -0500, Gabriel Dos Reis wrote:
| > On Fri, 26 Oct 2007, Jakub Jelinek wrote:
| > 
| > | On Fri, Oct 26, 2007 at 01:34:02PM -0500, Gabriel Dos Reis wrote:
| > | > No, one would need to set a `scope' cutoff.
| > | 
| > | So should reinit_cxx_pp set enclosing_scope to current_function_decl
| > | instead of 0?
| > 
| > Yes, try it out if it works.
| 
| With the additional patch below other/ptrmem8.C reports:
| /usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C: In function 'void foo(void (A::*)())':
| /usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C:9: error: invalid use of 'a.*f' to form a pointer-to-member-function
| /usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C:9: note:   a qualified-id is required
| /usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C: In function 'void bar(void (A::*)())':
| /usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C:15: error: invalid use of 'p->*f' to form a pointer-to-member-function
| /usr/src/gcc/gcc/testsuite/g++.dg/other/ptrmem8.C:15: note:   a qualified-id is required
| and no regressions appear in regression testing.
| Ok to commit both patches to the trunk?

Thanks!

-- Gaby