It not work when you render 汉字 by code tag

[lar-ry]

I want to render <code>汉字</code>,

it will display the tip:

invalid literal for int() with base 16: ''

v55.0 is normal, but v56.0b1, v56.0 and v56.1 is abnormal.

[liZe]

Hi!

Could you please send the PDF generated by version 55 and the one generated by version 56 (if the error doesn’t make WeasyPrint crash)?

[lar-ry]

version 55:
work normal.

Successfully installed weasyprint-55.0
PS C:\Users\larry\md2pdf> python .\md2pdf.py doc/index.md -s .\style\paper-zh.less -r all
🎉 doc/index has converted.
PS C:\Users\larry\md2pdf> 

generated pdf: index.pdf

---

version 56:
it work abnormal and can not make pdf successful.

Successfully installed weasyprint-56.0
PS C:\Users\larry\md2pdf> python .\md2pdf.py doc/index.md -s .\style\paper-zh.less -r all
invalid literal for int() with base 16: ''
PS C:\Users\larry\md2pdf>

[liZe]

I have two questions:

1. Could you please share (by mail) the font file used in your document with version 55? It’s 新宋体.
2. We need to have a traceback to know why WeasyPrint crashes. Could you please make your md2pdf.py script display this whole traceback?

[lar-ry]

1. It's SourceHanSansSC-Regular.otf and SourceHanSerifSC-Regular.otf:
   SourceHanSansSC-Regular.zip
   SourceHanSerifSC-Regular.zip
2. whole traceback:

   PS C:\Users\larry\md2pdf> python .\md2pdf.py doc/index.md -s .\style\paper-zh.less -r all
   Traceback (most recent call last):
   File "C:\Users\larry\md2pdf\md2pdf.py", line 138, in <module>
       convert(args.markdown, args.style, args.reserve)
   File "C:\Users\larry\md2pdf\md2pdf.py", line 58, in convert
       html.write_pdf(file_name + ".pdf")
   File "C:\Users\larry\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\weasyprint\__init__.py", line 201, in write_pdf
       .write_pdf(
   File "C:\Users\larry\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\weasyprint\document.py", line 335, in write_pdf
       pdf = generate_pdf(
   File "C:\Users\larry\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\weasyprint\pdf\__init__.py", line 445, in generate_pdf
       pdf_fonts = build_fonts_dictionary(pdf, fonts, optimize_size)
   File "C:\Users\larry\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\weasyprint\pdf\fonts.py", line 83, in build_fonts_dictionary
       _build_bitmap_font_dictionary(
   File "C:\Users\larry\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\weasyprint\pdf\fonts.py", line 208, in _build_bitmap_font_dictionary
       bits = bin(int(data.hex(), 16))[2:]
   ValueError: invalid literal for int() with base 16: ''
   PS C:\Users\larry\md2pdf>
   

[liZe]

Thanks a lot for this very useful answer.

The problem you have is caused by a bitmap font. Since version 56, WeasyPrint supports bitmap fonts, that’s why the rendering was working with version 55, using another font. Unfortunately, the font you shared is not bitmap, so it’s not the font causing the problem.

If you have some Python skills, it could be possible to save this font using a debugger (the content of the file is in font.file_content). Or maybe you know which bitmap fonts are installed on your system. Would you be interested in trying to find the font causing the problem?

[liZe]

@lar-ry Is there a way we could help you solving this problem?

[lar-ry]

I had to continue using version 55 because this font was required.

@lar-ry Is there a way we could help you solving this problem?

[liZe]

I had to continue using version 55 because this font was required.

OK, I understand.

The bug is actually caused by another font, a bitmap font installed on your system. The problem is that we don’t know which font it is.

Do you have some programming skills? If so, I could tell you how to find the font, so that I can fix the problem. But if you don’t have programming skills or don’t want to spend more time on this, no problem 😁️, we can then close this issue.

[lar-ry]

I want to try, what do I need to do?

[liZe]

I want to try, what do I need to do?

Thank you!

Here’s what you can do:

• Install WeasyPrint 56.1
• Open C:\Users\larry\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.10_qbz5n2kfra8p0\LocalCache\local-packages\Python310\site-packages\weasyprint\pdf\fonts.py
• Add open('C:\\Users\\larry\\font-' + font.hash, 'wb').write(font.file_content) on line 40. You should have something like:

    for font in fonts.values():
        open('C:\\Users\\larry\\font-' + font.hash, 'wb').write(font.file_content)
        widths = pydyf.Array()

• Generate your document (you’ll have the bug, it’s normal).

You should now have one or more files named font-XXXXXX (with uppercase characters instead of X) created in your C:\Users\larry\ folder. You can attach them in a comment or send them by mail, that’s the file I need!

[lar-ry]

fonts.zip
I did it, and it created 3 files in fonts.zip

[lar-ry]

Font files:

• Source Han Sans SC
• Source Han Serif SC
• 宋体

PS: You are really professional, thank you.

[liZe]

Thank you very much for your time, I can now reproduce your bug! I’ll come back as soon as I find a way to fix it.

[liZe]

The bug should now be fixed. The 宋体 font includes both bitmap characters and vector characters. We should use the vector variant (see #1736), removing the bug.

Moreover, the crash happened when characters have no data (ie. they have a width of 0). If we ever meet these characters in bitmap-only fonts again, we’ll now correctly include them.