LI74729: -QMBCS PRODUCES WRONG BYTECODE

June 2009 Update for XL C/C++ Advanced Edition for Linux, V9.0
February 2010 Update for XL C/C++ Advanced Edition for Linux, V9.0
February 2011 Update for XL C/C++ Advanced Edition for Linux, V9.0

APAR status

• Closed as program error.

Error description

• The C++ compiler generates bad unicode to wide character
  literals. C compiler works.
  C++ compiler fails for all versions.
  
  In the following test case, the unicode value passed to libc
  routine wcstombs is invalid.
  If correcting the value passed to wcstombs in dbx, test case
  returns correct output and values.
  
  
  $ cat main.c
  #include <stdio.h>
  #include <stdlib.h>
  
  int main( void )
  {
   char mbsbuf[5];
   char* plocale;
   size_t nReturn;
  
   plocale = setlocale( LC_ALL, "" );
   printf( "setlocale to %s\n", plocale );
  
   nReturn = wcstombs( mbsbuf, L"?", 5 );
   printf( "After wcstombs: %s, nReturn: %d\n", mbsbuf, nReturn );
  
          return 0;
  }
  
  COMMAND LINE:
  $ /usr/vacpp/bin/xlC main.c -qmbcs -+
  $ ./a.out
  
  EXPECTED OUTPUT:
  setlocale to Ja_JP Ja_JP Ja_JP Ja_JP Ja_JP Ja_JP
  After wcstombs: ?, nReturn: 1 <=HERE!!!!
  
  ACTUAL OUTPUT:
  setlocale to Ja_JP Ja_JP Ja_JP Ja_JP Ja_JP Ja_JP
  After wcstombs:  , nReturn: 2  <=HERE!!!!
  
  
  A reduced test case to demonstrate the problem:
  
  <file name="test_Ja_JP.c">
  #include <stdio.h>
  #include <wchar.h>
  wchar_t  test_wch  = L'&#65521;';  // Japanese character '&#65521;' may be
  altered by CMVC
  wchar_t *test_wstr = L"&#65521;";  // Japanese character '&#65521;' may be
  altered by CMVC
  int main()
  {
      printf("%x %x\n", (unsigned int)test_wch, (unsigned
  int)test_wstr[0]);
      return 0;
  }
  </file>
  
  COMMAND LINE:
  $ xlc++ -+ -qmbcs test_Ja_JP.c
  $ ./a.out
  
  EXPECTED OUTPUT:
  ff71 ff71
  
  ACTUAL OUTPUT:
  b1 b1
  

Local fix

• N/A
  

Problem summary

• USER AFFECTED:
  Programs which use wide character and string literals
  encoded in Japanese locale (Ja_JP) via -qmbcs.
  
  PROBLEMDESCRIPTION:
  Under -qmbcs, the bytecode produced from wide character
  or string literals is incorrect, if the literals are
  1-byte MBCS.
  

Problem conclusion

• Under -qmbcs, the compiler now converts wide char and string
  literals into the internal representation, using the mbtowc()
  system function, regardless whether the MBCS is 1-byte or
  multi-byte.
  

Temporary fix

Comments

APAR Information

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

  IZ51643

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

Fix information

• Fixed component name

  XL C/C++ RHEL5

• Fixed component ID

  5724S7310

Applicable component levels

• R900 PSN IZ51643

     UP06/09/13