On Mon, 5 Dec 2005, Martin Schröder wrote:
http://sarovar.org/download.php/846/pdftex-1.40.0-alpha-20051205.tar.bz2
I've tried to intergrate as many patches as possible: - 386 grow obj_tab and dest_names arrays dynamically
this 386 is not yet included, there were two conflicting patches. I will integrate it into a fresh 453 patch.
It's roughly tested: The manual still compiles.
As Pawe/l found, compiling samplepdf.tex leads to acroread and gs crashes. Reason 1: the first dictionary was written before the PDF-1.5 string. Compare to pdf_begin_obj: procedure pdf_begin_dict(i: integer; pdf_os: boolean); {begin a PDF dictionary object} begin check_pdfminorversion; {<--- this line was missing} pdf_os_prepare_obj(i, pdf_os); This will be repaired in the next patch. Now comes the interesting part: The reason 2 why Acroread and gs still crash (but not xpdf :-) ist this harmless looking object, 4 0 obj: Hello created by \pdfobj{Hello} \pdfrefobj \pdflastobj This "object" is self-standing, not referenced from anywhere. This seems to be completely legal in PDF-1.4. But when Acroread or gs find it within an object stream, and everything really looks ok AFAICT, both just choke. It works ok, when i simply write \pdfobj{(Hello)} instead of \pdfobj{Hello}. And no hint in the PDF Reference. My guess is that when an object stream is used, certain types are given to an object, e. g. <<...>> or [...] or (...) which all means something, but a simple Hello gives problems. Anybody any idea about this? What do with such a solitary Hello? Regards, Hartmut