I'm having a brief look on the new API and (may caused by beer) I can't understand why the filter has to install its state in a hash map.
I agree in that it is not needed to work wiht a pdf_hash variable if we can use a void pointer, so I changed both the API and the null filter to use a pointer to void. Now the init function for the filter receives a void** value, while the apply function for the filter receives the void* value (that may be NULL if the filter is not using any private state). Please note that the void pointer is disposed using pdf_dealloc in the pdf_stm_filter_destroy function, that is called by pdf_stm_destroy. PS: I will be a bit more busy this week but will try to get at least one or two filters ported. Nice, many thanks.
