Re: libstdc++/2832

Re: libstdc++/2832

[bkoz]

The following reply was made to PR libstdc++/2832; it has been noted by GNATS.

From: bkoz@gcc.gnu.org
To: bkoz@gcc.gnu.org, gcc-gnats@gcc.gnu.org, kenny.simpson@gs.com,
  theonetruekenny@yahoo.com
Cc:  
Subject: Re: libstdc++/2832
Date: 26 May 2001 09:21:32 -0000

 Synopsis: rdbuf methods incorrect
 
 State-Changed-From-To: feedback->closed
 State-Changed-By: bkoz
 State-Changed-When: Sat May 26 02:21:32 2001
 State-Changed-Why:
     Per request.
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=2832&database=gcc

Re: libstdc++/2832

[bkoz]

The following reply was made to PR libstdc++/2832; it has been noted by GNATS.

From: bkoz@gcc.gnu.org
To: bkoz@gcc.gnu.org, gcc-gnats@gcc.gnu.org, kenny.simpson@gs.com,
  theonetruekenny@yahoo.com
Cc:  
Subject: Re: libstdc++/2832
Date: 24 May 2001 23:05:54 -0000

 Synopsis: rdbuf methods incorrect
 
 State-Changed-From-To: analyzed->feedback
 State-Changed-By: bkoz
 State-Changed-When: Thu May 24 16:05:54 2001
 State-Changed-Why:
     Hi. I agree with your analysis, and have made the requisite changes in the iostreams code. Here's the patch:
     
     http://gcc.gnu.org/ml/gcc-patches/2001-05/msg01761.html
     
     Can you please verify that this solves your problem? I'll be checking it in to both trunk and branch in a few minutes.
     
     thanks again,
     benjamin
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=2832&database=gcc

RE: libstdc++/2832

[Simpson, Kenny]

The following reply was made to PR libstdc++/2832; it has been noted by GNATS.

From: "Simpson, Kenny" <Kenny.Simpson@gs.com>
To: "'Simpson, Kenny'" <Kenny.Simpson@gs.com>,
	"'bkoz@gcc.gnu.org'" <bkoz@gcc.gnu.org>, gcc-gnats@gcc.gnu.org,
	nobody@gcc.gnu.org, theonetruekenny@yahoo.com
Cc:  
Subject: RE: libstdc++/2832
Date: Wed, 23 May 2001 14:13:05 -0400

 Also, from the 1997 Draft at
 http://anubis.dkuug.dk/jtc1/sc22/open/n2356/lib-iostreams.html#lib.ostringst
 ream.members
 
 It claims that ostringstream::rdbuf() is never NULL because it ALWAYS
 returns the private buffer.
 
 But, from
 http://anubis.dkuug.dk/jtc1/sc22/open/n2356/lib-iostreams.html#lib.basic.ios
 .members
 It states that
 basic_streambuf<charT,traits>* basic_ios::rdbuf(
 basic_streambuf<charT,traits>* sb );
 has the post condition that sb == basic_ios::rdbuf()
 
 So, clearly basic_ios::rdbuf() and ostringstream::rdbuf() must be different.
 
 Also note that the iostreams implementation as shipped with 2.95.2 handles
 this correctly,
 so this should be considered a regression.
 
 -Kenny

RE: libstdc++/2832

[Simpson, Kenny]

The following reply was made to PR libstdc++/2832; it has been noted by GNATS.

From: "Simpson, Kenny" <Kenny.Simpson@gs.com>
To: "'bkoz@gcc.gnu.org'" <bkoz@gcc.gnu.org>, gcc-gnats@gcc.gnu.org,
	Kenny.Simpson@gs.com, nobody@gcc.gnu.org, theonetruekenny@yahoo.com
Cc:  
Subject: RE: libstdc++/2832
Date: Wed, 23 May 2001 10:32:05 -0400

 I don't have a copy of the Standard in front of me, but I do have
 _Standard C++ IOStreams and Locales_ by Langer and Kreft.
 
 p72 starts with:
 "rdbuf() SUBTLETIES
 Note the subtle difference between a concrete stream class's rdbuf()
 function and the
 inherited version of rdbuf(): The inherited rdbuf() function
 basic_ios<charT,
 traits>::rdbuf() returns the pointer to the stream buffer that is maintained
 by the
 base class.  The concrete stream class's rdbuf() function returns a pointer
 to its contained
 stream buffer object."
 
 It goes on to include a simple example similar to the one I submitted with
 the report:
 "
 basic_filebuf<char> buf;
 buf.open("in.txt");
 
 basic_ifstream<char> ifstr;          // ifstr by default contains a file
 buffer object
 ifstr.basic_ios<char>::rdbuf(&buf);  // here the file buffer pointer is
 replaced using the
                                      // hidden base class version of the
 rdbuf() function
 
 basic_streambuf<char>* bp;
 bp = ifstr.rdbuf();                  // redefined rdbuf() returns pointer to
 contained, but
                                      // unused buffer
 bp = ifstr.basic_ios<char>::rdbuf(); // base class version of rdbuf()
 function returns
                                      // pointer to newly assigned and
 actually used file buffer.
 "
 
 -Kenny
 
 -----Original Message-----
 From: bkoz@gcc.gnu.org [ mailto:bkoz@gcc.gnu.org ]
 Sent: Tuesday, May 22, 2001 8:28 PM
 To: bkoz@gcc.gnu.org; gcc-gnats@gcc.gnu.org; Kenny.Simpson@gs.com;
 nobody@gcc.gnu.org; theonetruekenny@yahoo.com
 Subject: Re: libstdc++/2832
 
 
 Synopsis: rdbuf methods incorrect
 
 Responsible-Changed-From-To: unassigned->bkoz
 Responsible-Changed-By: bkoz
 Responsible-Changed-When: Tue May 22 17:27:50 2001
 Responsible-Changed-Why:
     Responsible
 State-Changed-From-To: open->analyzed
 State-Changed-By: bkoz
 State-Changed-When: Tue May 22 17:27:50 2001
 State-Changed-Why:
     This seems to be two bugs:
     
     1) pointers returned by rdbuf() not the same
     
     2) [io]stringstream/[io]fstream dtors delete streambuf object.
     
     
     I don't see where, in the standard #1 is guaranteed. Perhaps you can
 point this out to me/
     
     I do agree that #2 is a problem, and am working on a fix now.
     
     thanks,
     benjamin
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=2832&database=gcc

Re: libstdc++/2832

[bkoz]

The following reply was made to PR libstdc++/2832; it has been noted by GNATS.

From: bkoz@gcc.gnu.org
To: bkoz@gcc.gnu.org, gcc-gnats@gcc.gnu.org, kenny.simpson@gs.com,
  nobody@gcc.gnu.org, theonetruekenny@yahoo.com
Cc:  
Subject: Re: libstdc++/2832
Date: 23 May 2001 00:27:50 -0000

 Synopsis: rdbuf methods incorrect
 
 Responsible-Changed-From-To: unassigned->bkoz
 Responsible-Changed-By: bkoz
 Responsible-Changed-When: Tue May 22 17:27:50 2001
 Responsible-Changed-Why:
     Responsible
 State-Changed-From-To: open->analyzed
 State-Changed-By: bkoz
 State-Changed-When: Tue May 22 17:27:50 2001
 State-Changed-Why:
     This seems to be two bugs:
     
     1) pointers returned by rdbuf() not the same
     
     2) [io]stringstream/[io]fstream dtors delete streambuf object.
     
     
     I don't see where, in the standard #1 is guaranteed. Perhaps you can point this out to me/
     
     I do agree that #2 is a problem, and am working on a fix now.
     
     thanks,
     benjamin
 
 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=2832&database=gcc