Looking at the gmemoryinputstream api, I wonder if the following would not make more sense:


GInputStream *g_memory_input_stream_new (void)
void          g_memory_input_stream_add_data (const void     *data,
                                              gssize          len,
                                              GDestroyNotify *destroy);

This would a) allow the input stream to be based on more than one chunk of memory
and b) allow to use memory that must be freed by other means than g_free (e.g. munmap or g_slice_free)