FastCGI vs. ICU RegexMatch

fastcgi developers,

i have a problem with FastCGI (C++) and RegexMatch of the icu library.

When i try to compile a regex, like the following one, the regex is'nt 
working correctly anymore. I know, this is not the ICU mailing list and 
i should not bother you with my regex problems, but this one is a bit 
strange in combination with FastCGI_Accept():


#include "unicode/unistr.h"
#include "unicode/regex.h"

#include <string>
#include <stdlib.h>
#include "fcgi_stdio.h"

using namespace std;

int main (int argc, char *argv[])
{
   RegexMatcher*   pRE;
   UErrorCode      status;
   pRE    = 0;
   status = U_ZERO_ERROR;

   int nOptions = UREGEX_UWORD|UREGEX_DOTALL;

   while(FCGI_Accept() >= 0)
   {
     pRE = new RegexMatcher("\\b(\\p{Letter}+?)\\b", nOptions, status);
     string sBuffer = "動ゲーム";

     string          sMatch;
     UnicodeString   usMatch;

     UnicodeString usBuffer = sBuffer.c_str();
     pRE->reset( usBuffer );
     while( pRE->find())
     {
       usMatch = pRE->group( 1, m_status );
       sMatch = bs_helper::BsHelper::GetStringFromUnicodeString(usMatch);
       cout << sMatch << endl;
     }
     delete pRE;

     printf("Content-type: text/html\r\n\r\n");
     (...)
   }
}

The regex should trench the words (there are two in the buffer "動" and 
"ゲーム") and output it with cout.

It works if i run it from console, without cgi.
It won't work if i run it through FastCGI. All environment variables are 
set correctly with -initial-env.

A bit more quaint is, it'll work through FastCGI if i change the code 
that way:

(...)
   int nOptions = UREGEX_UWORD|UREGEX_DOTALL;
   pRE = new RegexMatcher("\\b(\\p{Letter}+?)\\b", nOptions, status);

   while(FCGI_Accept() >= 0)
   {
     string sBuffer = "動ゲーム";
(...)

I use mod_fastcgi 2.4.2 for Apache 2. I compile my code on Debian with 
g++ 4.2.0

Regards

Thilo
___________________________________
fastcgi-developers mailing list
http://fastcgi.com/fastcgi-developers/