Edge displays “123456” in PDF but prints “114447”

[danso]

Just in case you don't click through, the bug submitter refers to a color PDF (not a scan, so image compression artifacts are not an issue) that is similar in appearance to a periodic table. That is, it is not the sequence of `123456` that is mistranslated into `114447`, but a sequence of 6 table cells, each containing a single digit.

It's not just the numbers that are misprinted, but the text inside those cells too, which suggests that Edge's PDF engine is re-rendering the original PDF, rather than printing the original PDF as is, which I thought was the entire point of using PDF in the first place.

But maybe this is an edge case? In the sense that Microsoft assumes that given a PDF file, if a user wants to "Print to PDF", the user should just save the PDF file. "Print to PDF" is ostensibly used to convert HTML/DOC into PDF format.

[saurik]

As someone who works in PDFs constantly (due to work in local government), I would say the point of a PDF is to be able to reproduce the result given a file, not to assume the file cannot be changed... you can easily edit and save PDF files using Acrobat, for example.

It is common that when you "print to PDF" you take the output of the printer and serialize that to PDF. I use this feature often on my Mac (which I think many would claim has excellent support for dealing with PDF files) to build a PDF that is stripped of any interactive forms: so as to get an output which is only the PDF "as printed".

[jahewson]

> which suggests that Edge's PDF engine is re-rendering the original PDF, rather than printing the original PDF as is

That's to be expected. The bitmap which Edge has rendered to the screen is not what will be sent to the printer driver. Instead, rich vector graphics will be sent. On Windows, the native print format is XPS, so this is most likely a bug in how Edge converts PDF to XPS for printing.

For simpler use cases Windows' graphics APIs can be used to both render to bitmap and to XPS but when printing something as rich and sophisticated as PDF better results are achieved by directly targeting the native print format, such as PCL, PostScript, or XPS. I suspect that's what the Edge devs have done and why it's producing different results on screen and in print.

[cesarb]

To make things even more interesting, the "original" PDF seems to have been generated by Ghostscript 8.15 and PScript5.dll 5.2, that is, it was also "printed to PDF" (from Microsoft Word, I presume).

[stephanheijl]

I have also encountered something similar whilst attempting to print a ticket to a major amusement park in Europe using Edge. The page the tickets were on secured by a login mechanism, and attempting to print the tickets resulted a page with an error. I had to save the PDF to the computer and print from there to get the proper output. It definitely seems like Edge re-renders or even re-requests the PDF before printing.

[dluc]

Nice pun, it's an edge case...

[givinguflac]

|Maybe this is an edge case?

Slow clap.

[type0]

> But maybe this is an edge case?

It is, Edge caused case

[kiliancs]

Definitely an Edge case.

[nom]

The PDF "format" never fails to amuse me. Check out the talk "OMG WTF PDF" [0] from the 27. Chaos Communication Congress, it's eye opening.

0: https://media.ccc.de/v/27c3-4221-en-omg_wtf_pdf

[tjalfi]

You would probably like the James Mickens video "Life As A Developer: My Code Does Not Work Because I Am A Victim Of Complex Societal Factors That Are Beyond My Control"[0]. He starts talking about the Adobe PDF reader at 19:30.

[0] https://vimeo.com/180568023

[H4CK3RM4N]

I still love that the SHA 1 collision was able to change the color of an image in a PDF, due to the junk data present.

[peapicker]

Thanks for that link.

[faragon]

This reminds me JBIG2 compression errors... [1]

[1] https://abbyy.technology/en:kb:tip:jbig2_compression_and_ocr

[bloaf]

Hence the joke in the bug report.

[dualogy]

Well.. 1+2+3+4+5+6 == 1+1+4+4+4+7 --- a bug with a sense for 'numerology'!

[EliRivers]

This reminds me of that photocopier that changed the numbers it was copying sometimes, through a dodgy image compression algorithm.

[teilo]

Yes, note the tongue-in-cheek reference reference to that problem in the article:

> (Possible workaround: Copy the document after printing using a Xerox copier.)

[ungzd]

Link: http://www.dkriesel.com/en/blog/2013/0802_xerox-workcentres_...

[cmurf]

The PDF goes through different rendering paths for display vs print. It's GDI+ for display, and WPF with an XPS spool file for print. So my guess is whatever does PDF to XPS filtering/conversion is getting something wrong; but then it could be complicated by an addtitional bug in the print driver which is why the report says the bug depends on what printer is used for printing.

[reacweb]

printing to pdf should be independent of printer.

[askvictor]

I've had similar issues with chrome's PDF viewer where it displays one number, but if I copy paste, it shows a different number.

[wdr1]

I wonder if it's related to when Xerox copiers changed numbers?

http://www.dkriesel.com/en/blog/2013/0802_xerox-workcentres_...?

[hoodoof]

x

[rileytg]

you're missing the joke

[steipete]

Maybe we really should bring https://pdfviewer.io to Windows. Looks like the default app is somewhat crap :P

[Sunset]

Why doesn't everyone use FoxitPDF in $current_year ?