release 1.23.0 not MacOS-arm64 compatible

[pjaol]

Please provide all mandatory information!

Using os x Ventura 13.4.1

poetry add pymypdf 

installs fine however when running you receive :

ImportError: dlopen(/Users/patrick/Projects/XXX/env/lib/python3.11/site-packages/fitz/_fitz.so, 0x0002): 
tried: '/Users/patrick/Projects/XXX/env/lib/python3.11/site-packages/fitz/_fitz.so' 
(mach-o file, but is an incompatible architecture (have 'x86_64', need 'arm64')), '/System/Volumes/Preboot/Cryptexes/OS/Users/patrick/Projects/XXX/env/lib/python3.11/site-packages/fitz/_fitz.so' (no such file), 
'/Users/patrick/Projects/XXX/env/lib/python3.11/site-packages/fitz/_fitz.so' (mach-o file, but is an incompatible architecture (have 'x86_64', need 'arm64')

backing down to

poetry add pymupdf==1.22.5

restores a working version

[julian-smith-artifex-com]

Thanks for reporting this.

Can you do pip install pymupdf instead?

In 1.23.0, the PyMuPDF wheel lists a new wheel PyMuPDFb as a requirement. Pip copes with this, installing both wheels automatically.

I'm not familiar with what poetry add does, perhaps it has an option to install such dependencies automatically? If not, you should be able to make things work by asking for both wheels explicitly, with something like poetry add pymypdfb pymypdf.

[Using two wheels in this way saves space on pypi.org, because we end up with just one PyMuPDFb wheel (containing the large MuPDF binaries) per OS.]

[bdonkey]

I've tried pip install pymypdf and have run into the same problems. Had to back down to pip install pymupdf==1.22.5.
Also tried installing both pymupdfb and pymupdf but same ARCH error.

[julian-smith-artifex-com]

This is odd, it works fine on other platforms. It looks like something may be installing a PyMuPDFb wheel for the wrong architecture. Or possibly we are not cross-compiling correctly when building macos-arm64 wheels.

If you could you do the following, it might allow us to figure out what's going on.

• Make sure your pip is up to date with python -m pip install -U pip.
• Clean up your venv with: pip uninstall PyMuPDF PuMuPDFb.
• Run pip install -vv PyMuPDF | tee out and attach the complete file out to this issue.

[JoshJarabek7]

Can confirm I am also having this issue.

Did a pip cache purge, deleted the venv, tried install fitz on its own, etc.

For pip install -vv PyMuPDF | tee out

Using pip 23.2.1 from /Users/josh/PARAKEET/BACK/.venv/lib/python3.11/site-packages/pip (python 3.11)
Non-user install because user site-packages disabled
Created temporary directory: /private/var/folders/wm/gl_n9_990fsdtz72dh5wm0bh0000gn/T/pip-build-tracker-_zpxdbl3
Initialized build tracking at /private/var/folders/wm/gl_n9_990fsdtz72dh5wm0bh0000gn/T/pip-build-tracker-_zpxdbl3
Created build tracker: /private/var/folders/wm/gl_n9_990fsdtz72dh5wm0bh0000gn/T/pip-build-tracker-_zpxdbl3
Entered build tracker: /private/var/folders/wm/gl_n9_990fsdtz72dh5wm0bh0000gn/T/pip-build-tracker-_zpxdbl3
Created temporary directory: /private/var/folders/wm/gl_n9_990fsdtz72dh5wm0bh0000gn/T/pip-install-xwc_x9ae
Created temporary directory: /private/var/folders/wm/gl_n9_990fsdtz72dh5wm0bh0000gn/T/pip-ephem-wheel-cache-8h1owpjr
Requirement already satisfied: PyMuPDF in ./.venv/lib/python3.11/site-packages (1.23.0)
Requirement already satisfied: PyMuPDFb==1.23.0 in ./.venv/lib/python3.11/site-packages (from PyMuPDF) (1.23.0)
Created temporary directory: /private/var/folders/wm/gl_n9_990fsdtz72dh5wm0bh0000gn/T/pip-unpack-itwxwvvf
Remote version of pip: 23.2.1
Local version of pip:  23.2.1
Was pip installed by pip? True
Removed build tracker: '/private/var/folders/wm/gl_n9_990fsdtz72dh5wm0bh0000gn/T/pip-build-tracker-_zpxdbl3'

[julian-smith-artifex-com]

It looks like some binaries within PyMuPDF 1.23.0-*macosx_11_0_arm64.whl wheels have been built incorrectly with a native x86_64 compiler instead of a cross compiler that targets arm64.

Apologies for this. We should be able to make a fixed release soon.

[julian-smith-artifex-com]

I've have a test wheel that may fix the problem.

The wheel is for MacOS, arm64 and Python-3.11.

It has been uploaded to test.pypi.org, and can be tested with:

pip uninstall --yes pymupdf
pip install -i https://test.pypi.org/simple pymupdf

Please let me know whether this works for you.

[julian-smith-artifex-com]

We've made a new release, PyMuPDF-1.23.1, which hopefully fixes this issue.

But we are currently unable to test MacOS arm64 wheels directly, so if people here could report how they get on with the new release, that would be very much appreciated.

[bdonkey]

Here is Claude summary of my testing:

Summary: Getting an architecture mismatch error when trying to import PyMuPDF on M1 Mac

Repro steps:

1. Created a new virtualenv (.venv) on my M1 Mac
2. Tried installing PyMuPDF via:
   1. pip uninstall --yes pymupdf
   2. pip install -i https://test.pypi.org/simple pymupdf
3. Also tried installing latest 1.23.1 release via pip install pymupdf
4. Tested with pymudf-utilities/OCR/tesseract2.py

Error:

Both installation methods produce the same import error:

Traceback (most recent call last):
  File "/Users/scottschmidt/scottdev/PyMuPDF-Utilities/_ssdev/OCR/tesseract2.py", line 22, in <module>
    import fitz
  File "/Users/scottschmidt/scottdev/PyMuPDF-Utilities/.venv/lib/python3.10/site-packages/fitz/__init__.py", line 22, in <module>
    from fitz.fitz import *
  File "/Users/scottschmidt/scottdev/PyMuPDF-Utilities/.venv/lib/python3.10/site-packages/fitz/fitz.py", line 14, in <module>
    from . import _fitz
ImportError: dlopen(/Users/scottschmidt/scottdev/PyMuPDF-Utilities/.venv/lib/python3.10/site-packages/fitz/_fitz.so, 0x0002): tried: '/Users/scottschmidt/scottdev/PyMuPDF-Utilities/.venv/lib/python3.10/site-packages/fitz/_fitz.so' (mach-o file, but is an incompatible architecture (have 'x86_64', need 'arm64')), '/System/Volumes/Preboot/Cryptexes/OS/Users/scottschmidt/scottdev/PyMuPDF-Utilities/.venv/lib/python3.10/site-packages/fitz/_fitz.so' (no such file), '/Users/scottschmidt/scottdev/PyMuPDF-Utilities/.venv/lib/python3.10/site-packages/fitz/_fitz.so' (mach-o file, but is an incompatible architecture (have 'x86_64', need 'arm64'))

Environment details:

• M1 Mac
• Python 3.10 virtualenv

Attaching pip install log and screenshot showing architecture mismatch.

pymmupdfPipTee.txt

Claude was better summarizing than my first attempt :)

[julian-smith-artifex-com]

Thanks for checking this, i'll carry on investigating what's going wrong.

[Aidenren1215]

I also encountered this issue, hopefully can solve it soon.

[julian-smith-artifex-com]

I've made some progress i think. This wheel works for me on a macos-arm64 gcc cfarm machine:

http://ghostscript.com/~julian/PyMuPDF-1.23.1-cp39-none-macosx_11_0_arm64.whl

(I only have a wheel for Python-3.9 at the moment, more on Monday.)

[julian-smith-artifex-com]

Here's a second test wheel now, for Python-3.11 on macos-arm64:

https://ghostscript.com/~julian/PyMuPDF-1.23.1-cp311-none-macosx_11_0_arm64.whl

(To test it, download to local directory and install with pip install PyMuPDF-1.23.1-cp311-none-macosx_11_0_arm64.whl.)

Note that currently i only have access to Python-3.9 on macos-arm64, so this wheel is untested.

[qcgm1978]

I meet some problems when import. At last install the latest pre-release version resolve it. pip install PyMuPDF==1.23.2rc1

[julian-smith-artifex-com]

Fixed in release 1.23.2.

[bdonkey]

Yes. Thank you. This seems to fix the ARCH error. However, when I run 1.23.2 against pymupdf-utilities/OCR/tesseract2.py I receive the error:

Traceback (most recent call last):
  File "/Users/scottschmidt/scottdev/PyMuPDF-Utilities/_ssdev/OCR/tesseract2.py", line 75, in <module>
    new_text = sb + get_tessocr(page, s["bbox"]) + sa
  File "/Users/scottschmidt/scottdev/PyMuPDF-Utilities/_ssdev/OCR/tesseract2.py", line 48, in get_tessocr
    ocrpdf = fitz.open("pdf", pix.pdfocr_tobytes())
  File "/Users/scottschmidt/scottdev/PyMuPDF-Utilities/.venv/lib/python3.10/site-packages/fitz/fitz.py", line 7414, in pdfocr_tobytes
    self.pdfocr_save(bio, compress=compress, language=language, tessdata=tessdata)
  File "/Users/scottschmidt/scottdev/PyMuPDF-Utilities/.venv/lib/python3.10/site-packages/fitz/fitz.py", line 7392, in pdfocr_save
    return _fitz.Pixmap_pdfocr_save(self, filename, compress, language, tessdata)
RuntimeError: No OCR support in this build

Does 1.23.2 not support OCR? I could run this file against 1.22.

[cbm755]

Not sure "+1"s are useful here, but this fixes my macOS arm64 builds, thanks for quick response.

Downstream reference: https://gitlab.com/plom/plom/-/merge_requests/2217/

[julian-smith-artifex-com]

However, when I run 1.23.2 against pymupdf-utilities/OCR/tesseract2.py I receive the error:

RuntimeError: No OCR support in this build

This is fixed in PyMuPDF-1.23.3.