Missing "Size" info in the trailer

[talcher]

Hi,

I use WeasyPrint to create PDF.
I recently upgraded my libraries to WeasyPrint 59.0 and pypdf 3.10.0.

My aim is to get the trailer of a PDF I created with WeasyPrint. In this trailer, I need the "Size" field, among others.

Previously, I got something like:

{'/Size': 14, '/Root': IndirectObject(3, 0, 140016423632752), '/Info': IndirectObject(2, 0, 140016423632752)}

Now, I get something like:

{'/Root': IndirectObject(3, 0, 140470815616880), '/Info': IndirectObject(2, 0, 140470815616880)}

If I force WeasyPrint to create a 1.4-versioned PDF, I get the trailer I expect.
However, this is not what I want. It would be much better not to force the 1.4 version.

Environment

• OS: Debian 11
• Python: 3.9.2
• WeasyPrint: 59.0
• PyPDF: 3.10.0

Code + PDF

from pypdf import PdfReader

reader = PdfReader("example.pdf")
trailer = reader.trailer.get_object()

Here are two PDF samples:

• no version forced : hello.pdf
• 1.4 version forced : hello_1.4.pdf

Suggestion

I have had a look at pypdf source code.
Correct me if I am wrong, but I think my issue is related to the "_read_xref_tables_and_trailers" function of the reader class, where trailer keys are defined that way: "trailer_keys = TK.ROOT, TK.ENCRYPT, TK.INFO, TK.ID"

If you need some more information, I would be glad to provide them.
Thank you for your help.

Alcher

[pubpub-zz]

Your analysis looks correct and I agree that TK.SIZE could be added to the list : the value is mandatory, as as part of the list, it will not be overwritten by old trailers.
Can you propose a PR and add a test to check the field is correctly extracted ?

[talcher]

Hello,
I just proposed a PR. Automatic checks have been successful.
Is that all you need? Sorry, I am a kind of a noob with GitHub workflows...

[MartinThoma]

Thank you for your input @talcher 🙏 Yes, this is all we need for the moment :-)