IV97566: UNEXPECTED ISTRINGSTREAM SETPRECISION FLOAT VALUE ROUNDING

XL C for AIX Fix Pack 9 (June 2017 Update) for 13.1
XL C/C++ for AIX Fix Pack 9 (June 2017 PTF) for 13.1
XL C for AIX Fix Pack 10 (February 2018 Update) for 13.1
XL C/C++ for AIX Fix Pack 10 (February 2018 PTF) for 13.1
XL C/C++ for AIX Fix Pack 11 (November 2018 PTF) for 13.1
XL C for AIX Fix Pack 11 (November 2018) for 13.1

APAR status

• Closed as program error.

Error description

• When storing a floating point value to an ostringstream object
  and retrieving it in an istringstream object, the resulting
  value has unexpected rounding compared to the setprecision()
  setting.
  
  
  
     === TEST CASE ===
  
  #include <iomanip>
  #include <iostream>
  #include <sstream>
  
  using namespace std;
  
  int main(int argc, char **argv)
  {
      float value      = 0.00342999981f;
      float checkValue = 0;
  
      cout.precision(17);
      cout << "original value:                                "
         << value << endl;
  
      ostringstream outstream_test;
      outstream_test << value;
      cout << "ostringstream string with no setprecision(9):  "
         << outstream_test.str().c_str() << endl;
  
      ostringstream outstream;
      outstream << std::setprecision(9) << value;
      cout << "ostringstream string after setprecision(9):    "
         << outstream.str().c_str() << endl;
  
      istringstream instream(outstream.str());
      instream >> std::setprecision(9) >> checkValue;
      cout << "instream string-to-float conversion:           "
         << checkValue << endl;
  
      if (value != checkValue)
         cout << "FAIL - Value does not match convert:" << endl;
      else
         cout << "PASS - Value matches convert:" << endl;
  
      cout << setw(55) << hex << *(unsigned int *)(&value) <<endl;
      cout << setw(55) << hex << *(unsigned int *)(&checkValue)
         << endl;
  
      return 0;
  }
  
  
  $ xlC test.cpp; ./a.out
  original value:                           0.0034299998078495264
  ostringstream string with no setprecision(9):
                                            0.00343
  ostringstream string after setprecision(9):
                                            0.00342999981
  instream string-to-float conversion:
                                            0.0034300000406801701
  FAIL - Value does not match convert:
                                            3b60c9d9
                                            3b60c9da
  $
  

Local fix

Problem summary

• PROBLEM DESCRIPTION:
  When converting an istringstream floating point string in the
  form of 0.<<two or more 0's>><<digits>> to binary, the XL C++
  Runtime changes the string to 0.0<<digits>> and calls strtof()
  with that string to convert to binary.  It then divides the
  result by 10 * <<number of 0's removed>>.  The extra rounding
  of the division operation results in precision loss.
  
  USERS AFFECTED:
  Users using an istringstream object containing floating point
  strings in the form of 0.<<two or more 0's>><<digits>>, e.g.,
  0.00342999981, and assigning the value to a 'float' variable.
  

Problem conclusion

• The runtime should not remove 0's from the string before
  conversion and then perform the division afterwards which
  results in precision loss.  The issue has been fixed.
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

  IV90071

• APAR is sysrouted TO one or more of the following:

Fix information

• Fixed component name

  XL C FOR AIX

• Fixed component ID

  5725C7100

Applicable component levels