Poppler issue: poppler-data should not be optional but a required dep

Batou · 2017-10-29 21:02:00

Hi all,

I've finally resolved an issue that's been bothering me for months. The issue was that some of the PDF files just wouldn't display text properly with any of the PDF viewers I tried (Evince, Xreader, Okular etc). Only MasterPDF Editor (under Arch) and viewers on other OSes would display text properly on all of the PDFs. After filing bugs with various viewers, I got a reply from GNOME/Evince developer and we managed to figure out what the issue is.

The issue was that I was missing the poppler-data package. Since Xreader/Evince/Okular, and almost every other FOSS PDF viewer, use poppler to render PDFs, they were all failing to render PDFs properly because I was missing poppler-data. Poppler-data package is described as "Encoding data for the poppler PDF rendering library" and it's an "Optional Dep" to poppler. It should be a hard dependency since, from my personal experience, 1 out of 5 PDFs will fail to render properly without it.

Evince/GNOME dev had this to say:

Evince is a shell to display documents with pages, for that purpose it uses
multiple backends, each one add supports for a file type.
Poppler provides support for rendering PDFs. I guess that Arch assumes the user
will install the backends they are interested to, but Arch developers might
know better.
Poppler-data could be a dependency for Poppler. I wonder when a user would not
want that.

Anyway, my suggestion is to make poppler-data a hard dependency and not an optional dependency to avoid all these display bugs.

Thanks.

V1del · 2017-10-29 21:12:12

The theoretical and individual necessity of a given package does not define if it has to be optional or not. The question is if the program runs without said package, which it does. This is even more of a niche case than many other "required" optional dependencies, since this will only really be relevant if your documents contain a lot of CJK encoded text. The info message during install states so quite plainly.

Batou · 2017-10-29 21:18:22

The thing is that I have documents/manuals made by a Japanese company which are completely in English. Yet they don't display English text at all without poppler-data. That's because they still encode them as Japanese for some reason. I've never had this issue with Debian-based distros.

Edit: What's even funnier, without poppler-data, CJK text displays fine in most of the documents I tried. But English text does not display in diagrams at all. See this pic of a manual without poppler-data: https://i.imgur.com/nuCe9yl.jpg

Edit 2: I also have electronic components data sheets that I downloaded from DigiKey/Mouser and they didn't display text properly either. Now they do. They were all in English but were produced by Chinese companies who typeset it under Chinese language locale yet they were in English. Go figure.

The description of what poppler-data does is clearly not complete and is misleading.

Last edited by Batou (2017-10-29 21:28:53)

Trilby · 2017-10-29 21:43:11

I've never had poppler-data installed on any of my systems and it's never caused me any issues, and I use poppler and poppler-based tools regularly.

I have no doubt you may need poppler-data for your pdfs, but it is definitely not required for much normal use of poppler. So optional dep sounds right to me.

Batou · 2017-10-29 21:56:08

Trilby wrote:

I've never had poppler-data installed on any of my systems and it's never caused me any issues, and I use poppler and poppler-based tools regularly.
I have no doubt you may need poppler-data for your pdfs, but it is definitely not required for much normal use of poppler. So optional dep sounds right to me.

Some of the display bugs are subtle. Sometimes you're missing a sentence or two and you don't even notice the text is missing. You've probably encountered this issue and never realized it's an issue.

There's a reason why it's a required dep and not optional one on other distros.

Anyway, I won't argue this issue any more. It's up to maintainer to make a decision. Just beware that without this package, your won't be seeing whole text in documents produced by large multinational companies... even when the text is not in CJK.

Last edited by Batou (2017-10-29 21:58:59)

fsckd · 2017-10-30 00:31:25

If the issue is as bad as you say, raise it on the bug tracker with examples or something concrete to convince the package maintainers that it should be a required dep. Discussing on the forums isn't likely to accomplish much, even if you have direct evidence of the problem, as many devs don't look on the forums for these kinds of threads.

Arch Linux

#1 2017-10-29 21:02:00

Poppler issue: poppler-data should not be optional but a required dep

#2 2017-10-29 21:12:12

Re: Poppler issue: poppler-data should not be optional but a required dep

#3 2017-10-29 21:18:22

Re: Poppler issue: poppler-data should not be optional but a required dep

#4 2017-10-29 21:43:11

Re: Poppler issue: poppler-data should not be optional but a required dep

#5 2017-10-29 21:56:08

Re: Poppler issue: poppler-data should not be optional but a required dep

#6 2017-10-30 00:31:25

Re: Poppler issue: poppler-data should not be optional but a required dep

Board footer