On Wednesday 18 April 2007 22:30, you wrote: > Since dvipdf does nothing else then > dvips -Ppdf -q -f "$infile" | > the difference must be again in the configuration of dvips. I guess the > problem is related to the fonts > lasxNNNN and laitNNNN > I have NO idea where they are coming from. Can you search your computer > with dlocate/locate and tell me where they are? Found the cause of part of the issue. Deleting $HOME/.texmf-var/ solved the checksum errors. The log now is: http://people.debian.org/~fjp/d-i_manual/ru.pdf2_sid.log > > As you can see the 'dvipdf' step now takes 8 _minutes_ instead of > > 20 _seconds_, and the 8 minutes is on a system that is a lot faster! > > The resulting PDF files are the same in both cases AFAICT. > > This is probably due to the generation of pixmap fonts, or some other > font generation. > > Can you run: > dvips -Ppdf "$infile" > and send me the output (of stdout and stderr?) Here it is: http://people.debian.org/~fjp/d-i_manual/ru.pdf2._sid.dvips.log > Does it *always* take so long time? Yes, but funnily it does not seem to be the dvips part of the process (see the timing in the log), but the piping into gs that takes most of the time! However, as gs (package gs-gpl) has not been uploaded since Okt 2006, the blame must still go to dvips which appears to somehow make things much harder for gs. > > - "Checksum mismatch" warnings for fonts that AFAICT are not even > > installed (nor available in Debian) These are gone now. > > - the "no %%Page comments generated" which I don't understand at all Still there, I also tried dvipdfm for the Russian PDF, and that resulted in: $ time dvipdfm -p a4 install.ru.dvi 2>/dev/null install.ru.dvi -> install.ru.pdf real 0m1.189s user 0m1.012s sys 0m0.112s Cheers, FJP
Attachment:
pgpCzUn8Iu4HX.pgp
Description: PGP signature