Arabic script for testing text extraction

habibi.pdf was generated using weasyprint 54.1-3 on debian unstable in July 2022, using the following command:

weasyprint habibi.html habibi.pdf

See also py-pdf/pypdf#1111

CMap Structure

habibi-oneline-cmap.pdf is the same file, but the beginbfchar stanza of the ToUnicode CMap is written with ASCII space delimiters between <srcString> <dstString> pairings, rather than newlines. That is, where habibi.pdf contains:

6 beginbfchar
<0003> <>
<03f2> <>
<0392> <>
<03f4> <>
<02f4> <>
<03a3> <062d064e0628064a0628064a0020>
endbfchar

habibi-oneline-cmap.pdf contains:

6 beginbfchar
<0003> <> <03f2> <> <0392> <> <03f4> <> <02f4> <> <03a3> <062d064e0628064a0628064a0020>
endbfchar

Otherwise the two files are exactly identical.

I believe text extraction should behave the same way on both files. From what i understand of the PDF specification, they are syntactically equivalent.