Re: [pdf-devel] Re: Bug in pdf_stm_flush()

[jemarch]

Hi Lukasz.

   pdf_stm_flush works fine, however i can't see any way of making
   more room in beckend.  Using pdf_stm_seek(stm, 0) cause that data
   in input bufer of null filter is lost.  I think making public
   pdf_stm_be_seek(stm->beckend, 0).  or ability to get beckend of
   writing stream and apply it to reading stream.

Streams operating on filesystem open files are always initialized with
a pdf_fsys_file_t that is known to the client. Thus, the client can
resize it, using the set_size function.

   Another idea is that pdf_stm_flush schould return PDF_EINVOP, when
   reading stream is used. Now there is returned 0.

Agreed.

   That is all. I am sending also patch with changes to:
   torture/unit/Makefile.am
   torture/unit/base/stm/tsuite-stm.c
   and a file torture/unit/base/stm/pdf-stm-flush.c

Thanks. I will apply the patch as soon as we get your papers. 

-- 
Jose E. Marchesi  <address@hidden>
                  http://www.jemarch.net
GNU Project       http://www.gnu.org