Some font thing failed when convert PDF to Image

[sidz]

Hi guys.

I have a PDF file which I'm trying to convert into the image(s) but for some reason vips shows some font thing failed message and removes some part of the text on the image.

Currently my code on PHP is looking like this:

$pdf = \Jcupitt\Vips\Image::newFromFile($filePath, ['access' => 'sequential']);

$pages = $pdf->get('n-pages');

for ($page = 0; $page < $pages; $page++) {
    $image = \Jcupitt\Vips\Image::newFromFile($filePath, ['page' => $page, 'access' => 'sequential']);

    $image = $image->thumbnail_image(1500);

    $image->writeToFile(sys_get_temp_dir() . "/pdf_{$page}.jpeg", ['Q' => 50]);
}

Guess this issue is not related to the php-vips bindings directly as I'm getting the same results when running:

$ VIPS_INFO=1 vips copy "file.pdf[n=-1,dpi=300]" file.jpg
VIPS-INFO: 09:27:52.335: found /usr/lib/vips-modules-8.15
VIPS-INFO: 09:27:52.335: using configure-time prefix
VIPS-INFO: 09:27:52.335: VIPS_PREFIX = /usr
VIPS-INFO: 09:27:52.335: VIPS_LIBDIR = /usr/lib
VIPS-INFO: 09:27:52.335: prefix = /usr
VIPS-INFO: 09:27:52.335: libdir = /usr/lib
VIPS-INFO: 09:27:52.336: searching "/usr/lib/vips-modules-8.15"
VIPS-INFO: 09:27:52.336: loading "/usr/lib/vips-modules-8.15/vips-jxl.so"
VIPS-INFO: 09:27:52.337: loading "/usr/lib/vips-modules-8.15/vips-poppler.so"
VIPS-INFO: 09:27:52.343: searching "/usr/lib/vips-plugins-8.15"
VIPS-INFO: 09:27:52.343: im_load_plugins: searching "/usr/lib/vips-8.15"

VIPS-INFO: 09:27:52.343: im_load_plugins: searching "/usr/lib"

VIPS-INFO: 09:27:52.350: threadpool completed with 2 workers
some font thing failed
some font thing failed
some font thing failed
some font thing failed
some font thing failed
some font thing failed
some font thing failed
some font thing failed
some font thing failed
some font thing failed
some font thing failed
some font thing failed
some font thing failed
some font thing failed
some font thing failed
VIPS-INFO: 09:27:52.659: threadpool completed with 4 workers

Is possible to convert pdf to image correctly?

[jcupitt]

Hello again,

some font thing failed

What a hilarious error message. That's not us, I'd guess something inside poppler has gone wrong, perhaps because this PDF contains an error or uses some strange feature. Did you make the PDF yourself? PDF export often has an "embed fonts" option, ticking this box can help.

I would try the same PDF in another poppler-based renderer, like xpdf or evince. If they can display it correctly, it's a libvips problem, if not (likely IMO) it's a poppler problem.

If it's poppler, your best best is to try PDFium, the alternative PDF backend on libvips. You'd need to be able to build libvips from source for this.

[jcupitt]

I should have said, PDFium is used for PDF rendering in Chrome, so a simple way to test compat there is to load your PDF into Chrome and see if it works. If you're on linux, evince is probably the best way to test poppler compatibility.

[kleisauke]

Alpine Linux is minimal and doesn't include fonts or font discovery tools by default. Perhaps installing a font package such as font-noto (which should also add fontconfig as a dependency) may help.

[sidz]

That's was it! Thank both of you @kleisauke, @jcupitt.

I can imagine that this is a hard work and could be annoying to reply to a such stupid questions. Your are amazing!

[sidz]

Did you make the PDF yourself?

No. It was generated by 3rd-party service: vokaal.com.

You'd need to be able to build libvips from source for this.

I've seen that PDFium has some issue with musl and that's the reason why Alpine team didn't bring it into their ecosystem yet. But this is not your (Vips team problem) :)

Alpine community request: https://gitlab.alpinelinux.org/alpine/aports/-/issues/16486
Gitub issue: bblanchon/pdfium-binaries#193 (comment)

I should have said, PDFium is used for PDF rendering in Chrome, so a simple way to test compat there is to load your PDF into Chrome and see if it works. If you're on linux, evince is probably the best way to test poppler compatibility.

Firefox could open that PDF correctly. Also I tried imagemagick + ghostscript and it converts PDF to image correctly.
Native Ubuntu PDF Viewer works correctly also.

I'm going to install fonts @kleisauke suggested above.

Just in case (if it gives you ideas): I didn't install any fonts when played with imagemagick + ghostscript.

[jcupitt]

@kleisauke is correct, I'd forgotten you were on alpine, you'll need to add all the font machinery. You'd think adding the poppler package would pull it in but I guess not.

[sidz]

I'd forgotten you were on alpine, you'll need to add all the font machinery.

No worries that's my fault. I didn't give you all details.

[sidz]

One more question related to particular this case: Is it possible to teach vips to throw an exception if something went wrong with poppler (for example) ?

The idea is pretty simple: I'd like to know when something went wrong instead of silently converting pdf to image with issues like I faced now

[jcupitt]

libvips has fail_on for this, eg:

$ vips pdfload
load PDF from file
usage:
   pdfload filename out [--option-name option-value ...]
where:
   filename     - Filename to load from, input gchararray
   out          - Output image, output VipsImage
optional arguments:
...
   fail-on      - Error level to fail on, input VipsFailOn
			default enum: none
			allowed enums: none, truncated, error, warning
...

But unfortunately poppler doesn't give out warning codes, so this won't do anything :( It'll work for formats like jpeg and tiff.

[sidz]

Got it @jcupitt. Thank you