Re: [pdf-devel] Streams solution proposal

[Juan Pedro Bolivar Puente]

>    Ok, I'll do it this evening.
> 
> Superb. Many thanks.
>

Done, patch attached. I also fixed a bug un pdf_stm_read that lead to an
infinite loop on small cache sizes (and deleted the buggy
pdf_stm_filter_set_be statment in pdf_filter_file_new).


There is another issue that I did not change but I think that should be
changed. pdf_flush (stm, PDF_TRUE) is called from pdf_stm_destroy. I
think that that behaviour is wrong, because twoo many ending characters
are printed if the user manually calls pdf_flush (stm, PDF_TRUE).

I think that we should make the finishing process totally transparent to
the user, as the user only knows about streams processed by filter
chaings, but should not care about filters details such as finishing
caracters. Instead, I think that we can automatically force a FINISH on
EOF. That would also allow printing the trailing characters when playing
on read mode -I think that the current behaviour of never finishing on
read mode is quite wrong).

What do you think? I can prepare that patch if you want.


> I am not sure about the convenience to use the pdf_filter utility to
> make unit tests. The unit tests for pdf_stm_read, pdf_stm_write, etc
> can load the data from files to make the comparisons.

Yep, you're right. I will do that when building the unit tests.

JP