Text Documents are altered when merged

[Phiwatec]

When using the PdfMerger to merge (append) two simple pdf files it merges them incorrectly.
The first is appended without a problem. But the second one is a mixture of both. If the pdfs are dissimilar this does not happen.
This happens both with pdfs created from LibreOffice Writer and an online convert Service.

$ python -m platform
Linux-5.10.0-13-amd64-x86_64-with-glibc2.31

$ python -c "import PyPDF2;print(PyPDF2.__version__)"
2.4.1

$ python --version
Python 3.9.2

This is a minimal, complete example that shows the issue:

#Basic example from docs
from PyPDF2 import PdfFileMerger, PdfFileReader
merger = PdfFileMerger()
merger.append("LO_first.pdf")
merger.append("LO_second.pdf")
merger.write("LO_out.pdf")
merger.close()

There a six attached files:
LO_first.pdf and LO_second.pdf are files created with Libreoffice Writer.
LO_out.pdf is the resulting wrong file
online_first.pdf and online_second.pdf are created using an online convert service from a plaintext file
online_out.pdf is the resulting wrong file

When using Firefox two view the resulting document it does not show a hidden character:

When using Chrome or Okular it shows a non printable character:

In both cases it should be bcde and not abc.

If the files contain very similar content ( "test1", "test2", "test3",etc.) the first file is used for all pages.

Thanks in advance for looking at this behavior.

[MartinThoma]

Thank you for reporting the issue. I'll have a look this week (and I hope @MasterOdin / @pubpub-zz have some time as well 😅)

[Phiwatec]

Thank you :)

[pubpub-zz]

I think the issue is due to the fact that the two initial pdf files are identifying a font named /F1 which has a different definitions in the two files. the first one only refers 3 characters "a" "b" "c" associated with coding 01 02 03 whereas in second find 01 02 03 04 are associated with "b" "c" "d" "e"
after merging the page 2 refers to the same font object as on page 1, which is inducing the incorrect data.
this is a bug confirmed

[MartinThoma]

@Phiwatec This issue was likely the same as #1062. It was introduced in 2.4.1 via #207. It was fixed in 2.4.2 (released moments ago) via #1063. A test was added to prevent this issue from happening again.

Could you please check if things work with PyPDF2==2.4.2 for you again?

[Phiwatec]

Thanks for the quick reply. It now works perfectly fine :)
Thank you for your time.