Using gtk+ from different threads is uneasy and error-prone ; the gdk lock must be obtained/released explicitly.

I won't deny that it's the programmer's job to protect the same gtk+ object to be accessed from different threads, but for the gdk lock, I think it should be gtk+'s job to ensure it takes the lock when it needs to.

In some sense, the program is gtk+'s client, not gdk's (nor xlib's), hence shouldn't have to care about the implementation details too much.