gtk window moved then crash - gcc

Question

I wanted to move my own gtk program from old MinGW to the latest one. I met a lot of questions in the past 2 weeks and couldn't find a way. Please refer to the other 2 questions I posted:
does mingw influence compilation a gtk program and compile gtk in mingw but failed.
Finally, I compiled gtk+2.24.10 by myself in MinGW. Followed are the details for environment:
GTK version gcc version
old mingw 2.10.11 -- from third party 3.4.5
new mingw 2.24.10 -- compiled by myself 4.7.2 or 4.6.2 -- I tried both
For the program in the old environment, everything runs well.
For the program in the new environment, the problem mentioned above "does mingw influence compilation a gtk program" disappeared. But there is a new problem I can't find any clue where it comes from:
When launching my program, if I move the GUI window, it crashed. I used gdb to debug, but if it crashed, system is frozen. The only way is to use "ctrl + alt + del" to activate task manager and stop the program. Then in GDB, where or bt full didn't show any useful information. If I don't move the GUI window, all the function in the window runs well. One more thing, my program uses openssl and pdflib which are from third party. The program links to the libs, but they are not used when the GUI window just is launched.
Is any bug in my own code? I don't think so because it runs well in the old MinGW environment. Some warnings about g_thread_create, g_thread_init is deprecated are in gtk+2.24.10 packages but don't show up in gtk+2.10.11 packages.
Is anything wrong with the gtk system, including gtk, glib, cario, atk etc., compiled by myself in MinGW? I can't tell. But gtk_demo included in gtk package runs well, I can move the program's window to wherever.
Is it related with gcc version? I can't tell. I googled and found someone complained that his gtk program complied by gcc 4.7.x is not stable and is affected by many sudden crashes. gcc 4.6.2 is absolutely suggested by someone. But in my case, it still crashed even compiled by gcc 4.6.2.
Is it related with openssl and pdflib which are from third party due to ABI issue? I can't tell.
M code is compiled in a Linux environment -- gtk+2.24.6. It does run well.
UPDATE 1:
I found some clue. The problem is from thread. But I still didn't know why it worked in the old MinGW but failed in the new MinGW. Followed is a simple example.
As I said, if I compile it in old MinGW, it worked -- the data, in fact a counter, is updated in the widget. And I can move the window.
If I compile it in new MinGW, it worked -- the data, in fact a counter, is updated in the widget. But if I move the window, it crashed. If I commented what are related with thread, then I can move the window, but data won't be updated in the window.
#include <gtk/gtk.h>
static int counter = 0;
static PangoLayout* layout;
static GdkGC* gc1;
static GdkGC* gc2;
//**/static GMutex* mu;
static gboolean on_expose_event(GtkWidget* widget, GdkEventExpose* event)
{
gchar the_string[20];
//**/g_mutex_lock(mu);
gdk_draw_rectangle(GDK_DRAWABLE(widget->window), gc1, TRUE, 0, 0, widget->allocation.width, widget->allocation.height);
snprintf(the_string, 20, "%d", counter);
pango_layout_set_text(layout, the_string, -1);
gdk_draw_layout(GDK_DRAWABLE(widget->window), gc2, 180, 120, layout);
//**/g_mutex_unlock(mu);
g_print (".");
return FALSE;
}
gpointer func(gpointer data)
{
//**/g_usleep(10000);
GdkWindow* window = GDK_WINDOW(data);
while(TRUE)
{
gdk_threads_enter();
gdk_window_invalidate_rect(window, NULL, FALSE);
gdk_threads_leave();
//**/gdk_window_process_updates(window, FALSE);
if(counter % 100 == 0) g_print("X");
g_usleep(10);
++counter;
}
return FALSE;
}
int main(int argc, char** argv)
{
GtkWidget* window;
GtkWidget* drawer;
GdkColormap* colormap;
GdkColor color;
g_thread_init(NULL);
gdk_threads_init();
gtk_init(&argc, &argv);
//:Create widgets and
window = gtk_window_new(GTK_WINDOW_TOPLEVEL);
drawer = gtk_drawing_area_new();
gtk_widget_set_size_request(drawer, 400, 300);
gtk_container_add(GTK_CONTAINER(window), drawer);
//.
gtk_widget_show_all(window);
//:Initializing Graphic Contexts
colormap = gtk_widget_get_colormap(GTK_WIDGET(drawer));
layout = gtk_widget_create_pango_layout(GTK_WIDGET(drawer), NULL);
gc1 = gdk_gc_new(GDK_DRAWABLE(GTK_WIDGET(drawer)->window));
gc2 = gdk_gc_new(GDK_DRAWABLE(GTK_WIDGET(drawer)->window));
gdk_color_parse("#000", &color);
gdk_colormap_alloc_color(colormap, &color, FALSE, TRUE);
gdk_gc_set_background(gc1, &color);
gdk_color_parse("#fff", &color);
gdk_colormap_alloc_color(colormap, &color, FALSE, TRUE);
gdk_gc_set_foreground(gc2, &color);
//.
g_signal_connect(G_OBJECT(window), "destroy", G_CALLBACK(gtk_main_quit), NULL);
g_signal_connect(G_OBJECT(drawer), "expose_event", G_CALLBACK(on_expose_event), NULL);
//**/mu = g_mutex_new();
//Run the problematic thread!
g_thread_create(func, GTK_WIDGET(drawer)->window, FALSE, NULL);
gdk_threads_enter();
gtk_main();
gdk_threads_leave();
//**/g_mutex_free(mu);
return 0;
}
Compile:
gcc -g -o example update_widget.c `pkg-config --libs --cflags gtk+-2.0 gthread-2.0`
UPDATE 2:
I checked config.log of compilation of glib and found the words below:
configure:26678: WARNING: I can't find the MACRO to enable thread safety on your
platform (normally it's _REENTRANT). I'll not use any flag on
compilation now, but then your programs might not work.
Please provide information on how it is done on your system.
Does it influence the thread function?
My configure for compilation of glib is:
export CFLAGS="-march=i686"
./configure --prefix=$HOME/gtk+$GTK_VERSION --with-threads=win32 --with-pcre=internal
make
make install