-2

My LaTex file has .pdf picture files in it. My own latex compiler handles them smoothly, but the ArXiv compiler refuses them, adding unintelligible error statements. Which statements does it accept? I was using

\includegraphics[height=100pt]{myfile.pdf} -> failure  (file has bounding box 440 250 )

I tried

\includegraphics[width=.75\textwidth,natwidth=440,natheight=250]{myfile.pdf} ->failure

No clue from ArXiv’s “help file”

3
  • 1
    it looks like you had configured arxiv to use latex not pdflatex (latex/dvips can not include PDF files) Jun 21, 2021 at 12:56
  • 1
    see arxiv.org/help/submit_tex#pdflatex Jun 21, 2021 at 12:57
  • Yeah, the arXiv compiler has some... let's call them "idiosyncrasies". Could you post the exact error messages (actually only the first one is usually relevant)?
    – campa
    Jun 21, 2021 at 13:08

2 Answers 2

1

The answer turns out to be simpler: ArXiv was lying to me. It said there was a lethal error in my graphicx pictures, but it just meant that it did something wrong in its first try to process my file, while the problem would have disappeared on its next try, but then another “lethal error” occurred, where it crashed for good. That lethal error happened when I had cut-and-pasted a reference to an article, written in some different editor. Perhaps some version of Word. It turns out some of the 256 characters are murderous to some versions of pdftex, not the one I am using but the latex version ArXiv is using. They are invisible. How could I have known? I think the designers of TeXShop should add a few program lines that weed out false characters. You can’t assume a user won't use invisible characters if you don’t even tell them not to.

6
  • if the error that you encountered is the one described at the arXiv texlive page regarding non-printing UTF-8 character encodings, you can sometimes get through them by adding the line: \UseRawInputEncoding as the first line of your tex file, as TeXLive 2020 has stricter inputs than 2016, by default.
    – Jake
    Aug 18, 2021 at 18:57
  • In using copy-and-paste from another source, apparently different characters are used to represent hyphens (-) and/or quotation symbols (` ‘ “), and maybe also spaces. The difference can only be seen with a microscope, if t all. I try to correct them, but sometimes I miss some. I don’t need software to remind myself, I am asking to add some lines of code in the latex software to detect, and as much as possible remove or replace, for other users who may be unaware at what virus has bitten them. The TexShop people were cooperative but couldn’t fix the problem as I wanted. Aug 20, 2021 at 13:18
  • The problem with detection is that they can be overly greedy and alter the meanings of some words. Because arXiv doesn't have staff that can reasonably scan the thousand+ submissions we receive daily, we rely on authors to ensure that what they submit means what they want it to mean. If you encounter errors again on your next submission, please feel free to reach out via our contact pages so that we can assist you at that time.
    – Jake
    Aug 21, 2021 at 14:08
  • My problem was that 1) the LaTex compiler of the ArXiv is more sensitive than the one I have myself, so that it became difficult to figure out what the new errors are. Actually the ArXiv was right in rejecting the false symbols that I had been using because they did generate minor typos in my own latex runs as well, which I had overlooked. But 2) the error messages were wrong and difficult to interpret (the errors were said to be in a figure, but they were in the symbols in the text) Aug 22, 2021 at 15:57
  • Of course having everything as intended is the responsibility of the user, and I would never ask the ArXiv to have staff checking the entries for mistakes by users, I am asking for some more lines in the software that can assist the user, making error messages easier to handle, but apparently this would also be difficult ... Aug 22, 2021 at 16:06
0

I had problems with specific characters (included in pdf figures) not showing up for an ArXiv submission. E.g. the letter "q" or the "%" sign.

What solved the issue was to add

\pdfinclusioncopyfonts=1

to the document preamble as described here. This article says the reason is that fonts with the same name (system name and name in embedded pdf font) are loaded only once and so some characters can go missing if these are not really the same font.

You must log in to answer this question.

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