Fonts not embedded in PDF

[slartibart70]

Hi,
i was using weasyprint 52.2 with python 3.8 on fedora32.
After upgrading to fedora33, my fonts are not embedded into the resulting pdf any more.
fedora33 uses python 3.9, i also upgraded to weasyprint 52.4... no change to the result.

What i did so far:

1. using weasyprint standalone (this is my main usage)

weasyprint in.html out.pdf

A pdf is generated, embedded fonts missing

2. using weasyprint as python script

from weasyprint import HTML, CSS
from weasyprint.fonts import FontConfiguration
font_config = FontConfiguration()
html=HTML('in.md.html')
html.write_pdf('out.pdf',font_config=font_config)

same result, no fonts in pdf

My Html uses an external css with @font-face
e.g.

@font-face {
  font-family: mySerif;
    font-weight: normal;
    font-style: normal;
    font-stretch: normal;
  src: local('Liberation Serif'),
       url(file:///usr/share/fonts/liberation-serif/LiberationSerif-Regular.ttf) format('truetype');
}

The browser has no problems rendering my embedded fonts, neither did weasyprint on fedora32/py3.8 (if that matters at all...)
So, my old working pdf output had these fonts (successfully embedded)

name                                 type              encoding         emb sub uni object ID
------------------------------------ ----------------- ---------------- --- --- --- ---------
AOAFFM+LiberationSerif               TrueType          WinAnsi          yes yes yes     36  0
VZUWYF+LiberationSerif               CID TrueType      Identity-H       yes yes yes     37  0
DIVSUD+RobotoMono-Regular            TrueType          WinAnsi          yes yes yes     84  0
JPSGLH+CairoFont-2-0                 Type 1C           WinAnsi          yes yes yes     85  0
HNYAUK+RobotoMono-Bold               TrueType          WinAnsi          yes yes yes    112  0
IQFZUN+RobotoMono-Regular            TrueType          WinAnsi          yes yes yes    113  0

The new one has only these default fonts:

name                                 type              encoding         emb sub uni object ID
------------------------------------ ----------------- ---------------- --- --- --- ---------
ANQKNJ+BitstreamVeraSans-Roman       TrueType          WinAnsi          yes yes yes     37  0
JBCTMJ+DejaVuSans                    CID TrueType      Identity-H       yes yes yes     38  0
XMKNHA+BitstreamVeraSans-Roman       CID TrueType      Identity-H       yes yes yes     83  0
YNMLOL+BitstreamVeraSans-Bold        TrueType          WinAnsi          yes yes yes     84  0
EZEEOY+BitstreamVeraSans-Oblique     TrueType          WinAnsi          yes yes yes    122  0

I had a (one!) font embedded into my pdf when using

weasyprint -v ${filename}.md.html ${filename}.md.pdf -s <(echo 'body { font-family: 'Liberation Serif' !important }')

with this font list:

name                                 type              encoding         emb sub uni object ID
------------------------------------ ----------------- ---------------- --- --- --- ---------
MNDGRP+LiberationSerif               TrueType          WinAnsi          yes yes yes     37  0
RLFUEA+LiberationSerif               CID TrueType      Identity-H       yes yes yes     38  0
AFSQCL+BitstreamVeraSans-Roman       TrueType          WinAnsi          yes yes yes     39  0
WKPJBS+LiberationSerif-Bold          TrueType          WinAnsi          yes yes yes     87  0
YTXGBQ+BitstreamVeraSans-Bold        TrueType          WinAnsi          yes yes yes    114  0
EZEEOY+BitstreamVeraSans-Oblique     TrueType          WinAnsi          yes yes yes    115  0

Overriding with user styelsheets and !important cannot be the solution...
I'm running out of ideas for now, any help?

[liZe]

I'm running out of ideas for now, any help?

Hi!

You can try to launch weasyprint with the --verbose option and see if WeasyPrint complains about fonts.

[slartibart70]

i did,
it's complaining about some css rules coming from pandoc (screen related), but nothing serious:

INFO: Step 1 - Fetching and parsing HTML - in.md.html
INFO: Step 2 - Fetching and parsing CSS - CSS string
INFO: Step 2 - Fetching and parsing CSS - CSS string
WARNING: Ignored `text-decoration: inherit` at 11:46, invalid value.
WARNING: Ignored `user-select: none` at 31:32, unknown property.
INFO: Step 2 - Fetching and parsing CSS - file:///home/user/.pandoc/pandoc4.css
INFO: Step 3 - Applying CSS
INFO: Step 4 - Creating formatting structure
INFO: Step 5 - Creating layout - Page 1
INFO: Step 5 - Creating layout - Page 2
INFO: Step 5 - Creating layout - Page 3
INFO: Step 5 - Creating layout - Page 4
INFO: Step 5 - Creating layout - Page 5
INFO: Step 5 - Creating layout - Page 6
INFO: Step 5 - Creating layout - Page 7
INFO: Step 5 - Creating layout - Page 8
INFO: Step 5 - Creating layout - Page 9
INFO: Step 5 - Creating layout - Page 10
INFO: Step 5 - Creating layout - Page 11
INFO: Step 5 - Creating layout - Page 12
INFO: Step 5 - Creating layout - Page 13
INFO: Step 5 - Creating layout - Page 14
INFO: Step 5 - Creating layout - Page 15
INFO: Step 5 - Creating layout - Page 16
INFO: Step 5 - Creating layout - Page 17
INFO: Step 5 - Creating layout - Page 18
INFO: Step 5 - Creating layout - Page 19
INFO: Step 6 - Drawing
INFO: Step 7 - Adding PDF metadata

[liZe]

Could you try to remove the local() part of your @font-face rules?

[slartibart70]

did that already... no change

[liZe]

Well… Could you share your HTML file, so that I can try to reproduce?

[slartibart70]

i made a stripped down version of the doc
I would expect the fonts to appear embedded in the pdf so that it looks like the html rendering in a browser
Thanks for your help!
test.zip

[liZe]

It works for me…

I really don’t know what could go wrong without a warning in the logs :/.

Another possible problem: do you have the latest version of tinycss2 (1.1.0)? Otherwise, url() with no quotes doesn’t work well.

[slartibart70]

nice to hear, but doesn't help me here :-)

I installed using pip, seems up-to date?
and i have put quotes around the urls.. still no change

$ pip list  --user
Package    Version
---------- -------
cairocffi  1.2.0
CairoSVG   2.5.2
cffi       1.14.5
cssselect2 0.4.1
pycairo    1.20.0
pycparser  2.20
Pyphen     0.10.0
tinycss2   1.1.0
WeasyPrint 52.4

[slartibart70]

and i tried on a completely different fedora33 machine, there weasyprint was installed as root (i know....)
pycairo 1.18.2
is the only difference

So... i assume this has something to do with fedora33 then?
But what exactly?

[liZe]

pycairo 1.18.2
is the only difference

WeasyPrint doesn’t use Pycairo (it uses CairoCFFI instead).

I installed using pip, seems up-to date?

It is.

So... i assume this has something to do with fedora33 then?
But what exactly?

OMG. I know. Fedora has updated Pango to 1.48.3 10 days ago, you need this fix: 47043a1.

We’ll release (again) another 52.x version, because it will cause a lot of issues.

Please let us know if it fixes your problem. Thanks a lot for the bug report, that’s really useful to know early that the new version of Pango has already been included in some distributions.

@FelixSchwarz: you may be interested in this issue. Backporting 47043a1 or upgrading WeasyPrint to the next 52.x version looks like a good idea.

(And for SemVer lovers: Pango 1.48.3 actually broke WeasyPrint. That’s not their fault, that’s not our fault.)

[FelixSchwarz]

@liZe Thanks for bringing this to my attention. I recently got an automated notification that WeasyPrint's test suite has even more test failures in Fedora but I did not have time to debug this.

[slartibart70]

you're my hero!!!
Thanks for pointing to the patch, appllied it and we're back to 'normal' again!!

[FelixSchwarz]

@slartibart70 if you are interested in helping out with Fedora's WeasyPrint, please let me know. ;-) Fedora uses very recent versions which means we spot problems (ideally) before/at the same time as upstream so fixing these is often upstream development. For example Fedora started to run test suites with Python 3.10 three months ago (and I have way too many bug reports about test failures with 3.10 waiting for me).

[slartibart70]

@FelixSchwarz
well, i use weasyprint for technical docs i write (i really don't want to use a wordprocessor for that). My workflow ist markdown/pandoc/weasyprint, so any errors coming up i'll notice immediately.
Or do you think of running further testsuites here on fc33?