system()-crash with phpGtk-win32

Hi, for several days now I'm fighting with the opportunity to start an binary exe-file on a Windows 32 system inside my phpGtk-script. The script is an Scanning-application which is used to scan and archieve images. To communicate between php and the TWAIN-compatible driver of my scanner (Agfa SnapScan e50) I use that binary exe-file named TwainCom.exe by Jungclaus Software (www.jse.de). This binary is a commandline-tool, which is able to use my Scanner-drivers functionality to scan images and save them in a given format such as JPEG or TIFF. To make it short: In my phpGtk-application the system()-call crashes my script. When I then kill my script (Not the TwainCom.exe-binary, which is still running), TwainCom.exe continues to run and finishes the job successful. If I'd instead kill the TwainCom.exe-binary, my phpGtk-script will continue to run. If I then run the system()-call a second time, everything works fine. The third try will crash my script again, I kill the TwainCom.exe-binary, my script continues, I try the fourth time, it works, on the fifth try it crashes again and so on and so on :( But: When I run the same system()-call just on command-line or inside a "normal" (small) php-script or phpGtk-script, it does _never_ crash, as it happens in the big phpGtk-script of mine. As I found out, there is a program named "Twunk_16", which is always getting loaded at the same time running the TwainCom.exe-binary. If this program is running (I am not able to start it manually, don't know why...), the system()-call inside my big phpGtk-script will work. But having "Twunk_16" running requires to start the TwainCom.exe-binary, wait till it crashes, then kill it. Killing it makes "Twunk_16" stay in memory until I run my system()-call. I also tried all this with the exec(), pasthru() and popen()/fwrite()-calls. They all have the same unstable behavior. I hope, you've got the point. I really need help on this, cause this project of mine is running out of time and it's gonna cost _alot_ ;( Please ask me _everything_, cause I'll be there... of course I may mail you a copy of my code, too. -- MfG, Sebastian G. : mailseppel@gmx.de : HH-germany :