preprocessor/9688: pasting "<" and "ValidStruct" does not give a valid preprocessing token

preprocessor/9688: pasting "<" and "ValidStruct" does not give a valid preprocessing token

[ediaz]

>Number:         9688
>Category:       preprocessor
>Synopsis:       pasting "<" and "ValidStruct" does not give a valid preprocessing token
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          rejects-legal
>Submitter-Id:   net
>Arrival-Date:   Thu Feb 13 09:46:02 UTC 2003
>Closed-Date:
>Last-Modified:
>Originator:     ediaz@veridis.com
>Release:        3.3 branch, pre-release.  2003-02-12
>Organization:
>Environment:
SuSE 7.3/Linux
>Description:
When I try to compile a file using macros of templates, I get a 'MyFileStruct.h:541:67: pasting "<" and "MyStruct" does not give a valid preprocessing token' message.

It worked perfectly with the 2.95.x serie and with the 3.2.1 version.

I've tryied to do it in two passes, with a --save-temps and a  g++ thefile.ii -c and it worked.  Like if the processed code was not the same as the preprocessor's output.

I've tried to make a dummy example which get that strange behaviour.

sh-2.05$ g++ test.cpp -o t1
test.cpp:57:32: pasting "<" and "Field" does not give a valid preprocessing token

Got a failure, but in two passes :

sh-2.05$ g++ test.cpp --save-temps
test.cpp:57:32: pasting "<" and "Field" does not give a valid preprocessing token
sh-2.05$ g++ test.ii -o t2        
sh-2.05$ 

Works fine.
>How-To-Repeat:
Just try the test.cpp file with the command lines in the description field.
>Fix:
A (IMHO dirty) workaround is to do it in two passes.
>Release-Note:
>Audit-Trail:
>Unformatted:
----gnatsweb-attachment----
Content-Type: text/plain; name="test.cpp"
Content-Disposition: inline; filename="test.cpp"

#include<stdlib.h>
#include<stdio.h>
#include<string.h>

template <class T>
class C
 {
  T hold;
  
  public:
   
    C(T to_hold)
     {
      hold=to_hold;
     }
    
    static void process(T a_value)
     {
      // NOP
      
      puts("Generic");
      
      return 0;
     }
 };

#define CM(x,va,vg)													\
template<>														\
class C < ##x >														\
{															\
public:															\
	static void process(x a_value)											\
	{														\
	 va														\
	 vg														\
	}														\
};

#define V(y) puts( #y );

struct Field
{
	char* name;
	char* value;
	
	bool operator==(const Field& o) const
	{
		return (strcmp(name,o.name)==0) && (strcmp(value,o.value)==0);
	}

	bool operator!=(const Field& o) const
	{
		return (strcmp(name,o.name)!=0) || (strcmp(value,o.value)!=0);
	}
};

CM( Field , V(name1) , V(name2));

int main(int argc,char* argv[])
 {
  Field f1;
  
  f1.name="afield";
  f1.name="avalue";
  
  C<Field>::process(f1);
 
  return 0;
 }

Re: preprocessor/9688: pasting "<" and "ValidStruct" does not give a valid preprocessing token

[Eric Diaz]

The following reply was made to PR preprocessor/9688; it has been noted by GNATS.

From: Eric Diaz <ediaz@veridis.com>
To: neil@gcc.gnu.org,
 ediaz@veridis.com,
 gcc-bugs@gcc.gnu.org,
 gcc-prs@gcc.gnu.org,
 nobody@gcc.gnu.org,
 gcc-gnats@gcc.gnu.org
Cc:  
Subject: Re: preprocessor/9688: pasting "<" and "ValidStruct" does not give a valid preprocessing token
Date: Thu, 13 Feb 2003 11:22:45 +0100

 On Thursday 13 February 2003 11:12, you wrote:
 > Synopsis: pasting "<" and "ValidStruct" does not give a valid preprocessing
 > token
 >
 > State-Changed-From-To: open->closed
 > State-Changed-By: neil
 > State-Changed-When: Thu Feb 13 10:12:26 2003
 > State-Changed-Why:
 >     Not a bug.  Two passes are different coz only one is doing macro
 > expansion!
 
 doh ... I had forgotten that ...
 
 Thank you very much.

Re: preprocessor/9688: pasting "<" and "ValidStruct" does not give a valid preprocessing token

[neil]

Synopsis: pasting "<" and "ValidStruct" does not give a valid preprocessing token

State-Changed-From-To: open->closed
State-Changed-By: neil
State-Changed-When: Thu Feb 13 10:12:26 2003
State-Changed-Why:
    Not a bug.  Two passes are different coz only one is doing macro expansion!
    
    Remove the ##.  And no, 3.2.1 was no different, but just a warning.
    
    2.95 accepted invalid constructs.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=9688