PDF with trailing null characters after EOF raises MalformedPDFError

[raldred]

I've got a PDF which a customer uploaded which looks to have been created by Atos using Fonet.
It appears to have a bunch of null characters after %%EOF so it raises

PDF::Reader::MalformedPDFError (PDF does not contain EOF marker)

Is there anyway to support this sort of file?

[yob]

What version of pdf-reader are you using? In theory, from v2.6.0 we can find the EOF with up to 5k trailing bytes (see #382), and prior to that we looked in the last 1024.

Maybe NULL bytes behave differently with the logic we have though?

[raldred]

2.15.1
I was just experimenting with the TRAILING_BYTECOUNT
Turns out that the EOF in the problamic file is at 5995

io.seek(-5995, IO::SEEK_END)
data = io.read(5995)
=> "\r\n%%EOF\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00

What are the downsides of increasing the TRAILING_BYTECOUNT further?

This file is only 15K overall!

[yob]

I wonder if there'd be a performance impact of making it ever bigger? I'm not opposed to increasing it though, and if we find a performance impact we could consider first doing a check for the common case of no trailing junk bytes