Craig> I have now updated from CVS, applied this patch, and Craig> recompiled. Craig> Individual tests work! Craig> Package updates via PUI work!!! (better that they used to) Craig> Stress test: Craig> (dotimes (i 200 "grep test passed") Craig> (unless Craig> (zerop (call-process "grep" nil '(t nil) nil Craig> "drivers" "c:\winnt\system.ini")) Craig> (error "Incorrect call-process return value!"))) Craig> runs to completion without showing the Blue Screen of Death!!!!! Craig> It first glance, I would say that we have a winner. That's great news! Craig> Craig Craig> P.S. When I quit XEmacs (c-X c-C) the following was printed Craig> to STDERR: Craig> Fatal error: assertion failed, file console.h, line 419, Craig> RECORD_TYPEP (obj, &lrecord_console) I assume this is related to another os_subtype check in ntproc.c: static BOOL CALLBACK find_child_console (HWND hwnd, child_process * cp) { DWORD thread_id; DWORD process_id; thread_id = GetWindowThreadProcessId (hwnd, &process_id); if (process_id == cp->procinfo.dwProcessId) { char window_class[32]; GetClassName (hwnd, window_class, sizeof (window_class)); if (strcmp (window_class, (os_subtype == OS_WIN95) ? "tty" : "ConsoleWindowClass") == 0) { cp->hwnd = hwnd; return FALSE; } } /* keep looking */ return TRUE; } I don't have a solution yet. What do you think we should do to determine os_subtype under mingw32? The sample program you sent looks equivalent to what src/ntheap.c has. Perhaps Andy can comment what to do? Anyway, I feel a lot better now than I did a few days ago. Best regards, Adrian Craig> and the shell hung waiting for XEmacs to exit. Craig> Otherwise nothing was printed to STDERR!