Hi folks, I'm currently working on a feature to perform OCR on PDF files that contain embedded images. We convert the PDF to an image using the

pdf2image

library and then run OCR on the image with

pytesseract

. While writing a unit test for this flow, I encountered an issue when running the test with Pants. Pants is unable to open the temporary image we get when converting a pdf -> image, and I’m seeing the error attached in the thread. Interestingly, the test runs fine with Poetry. I suspect this might be related to how Pants sandboxes tests, potentially restricting access to certain temp files generated during execution. Has anyone else run into this issue or have any suggestions on how to resolve it? Thanks!

Sorry for the trouble. I'm not sure about the specifics, but, in general, the best way to debug this sort of thing is to jump into the sandbox and execute code there: https://www.pantsbuild.org/stable/docs/using-pants/troubleshooting-common-issues#debug-tip-inspect-the-sandbox-with---keep-sandboxes In particular if you pass

--keep-sandboxes=on_failure

you'll see the failing test's sandbox printed, and then you can go to that directory and run the

__run.sh

... hopefully that reproduces the problem! Once you've reproduced it outside of a

pants ...

invocation, you can then debug it like "normal" e.g.

pytest --pdb ..

or whatever you need to do

Hmm, i think the issue happening is that whenever I convert the pdf to an image using the

convert_from_path

function in

pdf2image

, it temporarily stores the image in a /tmp folder on my computer. Unfortunately since pants creates a sandbox, it probably doesn't have access to this created /tmp folder as it doesn't exist in the sandbox..

That doesn’t sound quite right. The sandbox isn’t very strict, and absolute paths like

/tmp

should work fine (we use them in my work repo).