stdcxx-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Geoffrey Winn" <>
Subject Re: Problem building Tuscany SDO with stdcxx on Linux
Date Mon, 16 Oct 2006 16:17:07 GMT
Thank you both. Once I added -libsupc++ to the link, the the test program
builds cleanly and I can run it.

The good news is that from 111 test cases, 109 run successfully.

The bad news is that 2 fail in calls to the C++ library and then the test
program itself crashes - ironically, while printing the number of successful
tests :-)

I'd like to check the final crash first because I think it might be the
easiest to isolate. It happens when executing lines like

cout << "Tests passed: ";

This is in fact the first reference to cout in the program, so by way of an
experiment I inserted

cout << "Starting tests";

at the beginning and it crashes there instead. Does anyone recognise this

gdb prints the following, rather cryptic, backtrace

(gdb) bt
#0  0x00e92cdf in raise () from /lib/tls/
#1  0x00e944e5 in abort () from /lib/tls/
#2  0x00246592 in __rw::__rw_assert_fail (expr=0x2b3af3 "this->_C_is_valid
    func=0x2b3e60 "std::basic_streambuf<_CharT, _Traits>*
std::basic_filebuf<_CharT, _Traits>::setbuf(_CharT*, int) [with _CharT =
char, _Traits = std::char_traits<char>]") at
#3  0x0026d962 in std::basic_filebuf<char, std::char_traits<char> >::setbuf
(this=0xbfffa4d0, __buf=0x0, __len=512) at
#4  0x08073005 in sdotest::testUtil () at fstream:109
#5  0x0808ad51 in main (argc=1, argv=0xbfffa664) at main.cpp:106

In the case where the reference to cout is at the beginning of the program
then the source is ...

#include <stdio.h>

#pragma warning(disable:4786)

#include <iostream>
using namespace std;

#include "sdotest.h"

extern "C"{

#define TEST(testname)\
    value = testname;\
    if (value == 0) {\
        cout << "Test Failed: " << totaltests  << " " << #testname
<< endl;\
    testspassed += value;\

int main (int argc, char** argv)
    cout << "Starting SDO tests ...";      <=== Fails here.

The thing that puzzles me is that this is fairly unremarkable code compared
to all the other stuff that's workin OK.



On 12/10/06, Liviu Nicoara <> wrote:
> Martin Sebor wrote:
> > Geoffrey Winn wrote:
> >> Is the reference to __cxa_pure_virtual familiar to anyone?
> > (When using g++, the runtime is included in libstdc++).
> I stand corrected :-), g++ does not link libsupc++ anymore - these
> symbols are indeed in libstdc++. I do remember a time when downstream
> products on top of our stdlib were linking libsupc++ explicitly.
> Liviu

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message