std::ostringstream bizarre behaviour. Please help!

[Simon Pryor]

I am having some strange problems using std::ostringstream.

The simple stuff works okay, but trying to use:
std::ostringstream::str(const std::string&)
or:
std::ostringstream::ostringstream(const std::string&)

Gives some weird results on both Solaris & Linux.

Either that or I'm missing something. I've made a simple program to
demonstrate the problem:

-------------cut-as-ostringstreambug.cc---------------------------
#include <string>
#include <iostream>
#include <sstream>
int main(void)
{
int rc = 0;
std::ostringstream oss1;
oss1 << "foo";
oss1.str(std::string());
if (oss1.str() != std::string(""))
{ ++rc; std::cout << __LINE__ << ": Got \""
<< oss1.str() << '\"' << std::endl; }
std::ostringstream oss2(std::string("foo"));
oss2 << "bar";
if (oss2.str() != std::string("foobar"))
{ ++rc; std::cout << __LINE__ << ": Got \""
<< oss2.str() << '\"' << std::endl; }
std::ostringstream oss3;
oss3 << "foo";
oss3.str(std::string("bar"));
oss3 << "foo";
if (oss3.str() != std::string("barfoo"))
{ ++rc; std::cout << __LINE__ << ": Got \""
<< oss3.str() << '\"' << std::endl; }
return rc;
}
--------------------cut-end----------------------------------------

On Solaris:
CC -V
CC: WorkShop Compilers 5.0 98/12/15 C++ 5.0
CC -o ostringstreambug ostringstreambug.cc
ostringstreambug;echo $?
11: Got ""
16: Got "bar"
23: Got "foo"
3

On Linux:
g++ -v
gcc version 3.2.2
g++ -o ostringstreambug ostringstreambug.cc
ostringstreambug;echo $?
16: Got "bar"
23: Got "foo"
2

Now, I would expect to see no output and a return value of 0 on all
machines.

The only thing I can think of is that there is some sort of "freezing" the
way there was with the old "char* ostrstream::str()" stuff, but I don't
find anything that documents the intended behaviour.

Can anyone help?????????? please????

--

+------------------------------------------------------------------
| Simon Pryor, mailto:si***************@alcatel.be
| Alcatel Bell Space N.V.
| Berkenrodelei 33, phone: [phone]
| B-2660 Hoboken, Belgium. fax: [phone]
+------------------------------------------------------------------

[Peter Kragh]

Simon Pryor wrote:

#include <string>
#include <iostream>
#include <sstream>
int main(void)
{
int rc = 0;
std::ostringstream oss1;
oss1 << "foo";
oss1.str(std::string());
if (oss1.str() != std::string(""))
{ ++rc; std::cout << __LINE__ << ": Got \""
<< oss1.str() << '\"' << std::endl; }
std::ostringstream oss2(std::string("foo"));
Try this:

std::ostringstream oss2(std::string("foo"), std::ios_base::out |
std::ios_base::app);
oss2 << "bar";
if (oss2.str() != std::string("foobar"))
{ ++rc; std::cout << __LINE__ << ": Got \""
<< oss2.str() << '\"' << std::endl; }
std::ostringstream oss3;
Try:

std::ostringstream oss3(std::ios_base::out | std::ios_base::app);

oss3 << "foo";
oss3.str(std::string("bar"));
oss3 << "foo";
if (oss3.str() != std::string("barfoo"))
{ ++rc; std::cout << __LINE__ << ": Got \""
<< oss3.str() << '\"' << std::endl; }
return rc;
}
--------------------cut-end----------------------------------------

On Solaris:
CC -V
CC: WorkShop Compilers 5.0 98/12/15 C++ 5.0
CC -o ostringstreambug ostringstreambug.cc
ostringstreambug;echo $?
11: Got ""
16: Got "bar"
23: Got "foo"
3

On Linux:
g++ -v
gcc version 3.2.2
g++ -o ostringstreambug ostringstreambug.cc
ostringstreambug;echo $?
16: Got "bar"
23: Got "foo"
2

Now, I would expect to see no output and a return value of 0 on all
machines.

The only thing I can think of is that there is some sort of "freezing" the
way there was with the old "char* ostrstream::str()" stuff, but I don't
find anything that documents the intended behaviour.

Can anyone help?????????? please????

According to

http://www.roguewave.com/support/doc...uf.html#idx403

, you need to use out | app for the rdbuf. Otherwise the output sequence
points to the first character in the buffer.

I tried this on gcc version 3.3.1 and it worked. However, I also tried
with Microsoft's C++ compiler version 13.10.3052 and it didn't work
(maybe I should update it :-) ).

HTH.

BR,
Peter

[Simon Pryor]

Thanks for your help!

This fully fixed it for Linux and partially fixed it for Solaris.

However, when resetting to a NULL string, on Solaris, there was still a
problem (If you refer to my original posting, you can see there was already
a difference in output between Solaris & Linux.
Running Purify on Solaris also gave all sorts of FMR & UMR errors, with this:

std::ostringstream oss1(std::ios_base::out | std::ios_base::app);
oss1 << "foo";
oss1.str(std::string()); // or oss1.str("");
if (oss1.str().size())
{ ++rc; std::cout << __LINE__ << ": Got "
<< oss1.str().size() << "/" << int(oss1.str().data()[0])
<< std::endl; }

On Solaris, I get:

11: Got 1/0
1

So there is an internal bug in the Solaris STD library of C++ 5.0!

Maybe a workaround is always to "reset" to a space or newline or
something like that, though that's not always possible to guarantee
everyone doing this :-(

--

+------------------------------------------------------------------
| Simon Pryor, mailto:si***************@alcatel.be
| Alcatel Bell Space N.V.
| Berkenrodelei 33, phone: [phone]
| B-2660 Hoboken, Belgium. fax: [phone]
+------------------------------------------------------------------

[Moskvichev Nikolay]

Peter Kragh wrote:

According to

http://www.roguewave.com/support/doc...uf.html#idx403

, you need to use out | app for the rdbuf. Otherwise the output sequence
points to the first character in the buffer.

I tried this on gcc version 3.3.1 and it worked. However, I also tried
with Microsoft's C++ compiler version 13.10.3052 and it didn't work
(maybe I should update it :-) ).

It didn't work on Microsoft C++ compiler version 13.10.3077 and
Borland C++ compiler version 5.6.4 too. (Dinkumware & STLPort IIRC)

[Stephen Howe]

> Try this:

std::ostringstream oss2(std::string("foo"), std::ios_base::out |
std::ios_base::app);

Unfortunately there is nothing in the standard which indicates ostringstream
takes any notice of flags like std::ios_base::app. So implementations are
free to ignore them in constructors

Stephen

[Peter Kragh]

Stephen Howe wrote:

Unfortunately there is nothing in the standard which indicates ostringstream
takes any notice of flags like std::ios_base::app. So implementations are
free to ignore them in constructors

Maybe, I misunderstood chapter 27.7.3.1 (which I should have refered to
in the first place) then:

<quote>
Constructs an object of class basic_ostringstream, initializing the base
class with basic_ostream(& sb) and initializing sb with
basic_stringbuf<charT,traits,Allocator>( which | ios_base::out))
</quote>

Please correct me, if I'm wrong.

- Peter