2

I build a PDF document from a LaTeX source file as part of static Dwebsite generation, but every run of pdflatex outputs a file that hashes to a different IFPS Content ID. This triggers spurious regeneration of the root CID. Can pdflatex generate a PDF that is binary-equivalent to previous runs (assuming an unchanged input file) so the PDF output hashes to the same CID?

The pdflatex manpage doesn't list any options that appear relevant. Per http://theoval.cmp.uea.ac.uk/~nlct/latex/pdfdoc/pdfdoc/pdfdoc.html, I have added something like this to the top of the LaTeX file:

\pdfinfo{
   /Producer (pdfTeX)
   /Author ()
   /Title  ()
   /CreationDate (D:20040502195600)
   /ModDate (D:20040502195600)
   /PTEX.Fullbanner (Generated with pdfTeX)
}

Comparing the files in hexdiff shows deltas in an /ID property that looks like a GUID. This property doesn't appear to be modifiable by the \pdfinfo block. Is there any way to hardcode or disable the generation of this property?

1 Answer 1

5

You can set reproducible behaviour by setting the environment variables SOURCE_DATE_EPOCH and FORCE_SOURCE_DATE see texdoc pdftex section 4.

SOURCE_DATE_EPOCH should be the number of seconds since 1970 that you want to fix the date for, and FORCE_SOURCE_DATE should be 1

You can also control these fields from within the tex document eg \pdftrailerid controls the id at the end and \pdfsuppressptexinfo=-1 would suppress the printing of the info information.

2
  • I wasn't able to get \pdfsuppressptexinfo=-1 working, but a combination of \pdftrailerid{} and the \pdfinfo block I mentioned in my question seems to have done the trick. Thanks!
    – cqcallaw
    Oct 25, 2020 at 0:06
  • When validating my solution, I got tripped up thinking \pdfsuppressptexinfo=-1 was working, but it turned out to be a clock resolution issue. I was generating both PDFs with a chain of shell commands, and the mod/creation date format didn't have sufficient resolution to measure the delta. The solution: insert a sleep between invocations of pdflatex to make sure the difference in mod times is observable.
    – cqcallaw
    Oct 25, 2020 at 0:09

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .