py27-tensorflow1: ERROR: tensorflow-1.15.3-cp27-cp27m-macosx_10_13_x86_64.whl is not a supported wheel on this platform.

[ryandesign (Ryan Schmidt)]

​https://build.macports.org/builders/ports-10.12_x86_64-builder/builds/122650/steps/install-port/logs/stdio

ERROR: tensorflow-1.15.3-cp27-cp27m-macosx_10_13_x86_64.whl is not a supported wheel on this platform.

[ryandesign (Ryan Schmidt)]

Current logs for
​py27-tensorflow1 @1.15.4
​py36-tensorflow1 @1.15.4
​py37-tensorflow1 @1.15.4
​py38-tensorflow1 @1.15.4

The problem seems to be that although we are on macOS 10.12 it is compiling with -mmacosx-version-min=10.13 (which it should not do) which ultimately leads to:

[WARNING] This wheel needs a higher macOS version than the version your Python interpreter is compiled against.  To silence this warning, set MACOSX_DEPLOYMENT_TARGET to at least 10_13 or recreate these files with lower MACOSX_DEPLOYMENT_TARGET:  

  File "setup.py", line 306, in <module>
    keywords='tensorflow tensor machine learning',
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/setuptools/__init__.py", line 153, in setup
    return distutils.core.setup(**attrs)
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/distutils/core.py", line 148, in setup
    dist.run_commands()
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/distutils/dist.py", line 966, in run_commands
    self.run_command(cmd)
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/distutils/dist.py", line 985, in run_command
    cmd_obj.run()
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/wheel/bdist_wheel.py", line 328, in run
    impl_tag, abi_tag, plat_tag = self.get_tag()
  File "/opt/local/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/wheel/bdist_wheel.py", line 278, in get_tag
    assert tag in supported_tags, "would build wheel with unsupported tag {}".format(tag)
AssertionError: would build wheel with unsupported tag ('cp37', 'cp37m', 'macosx_10_13_x86_64')

We also see in the log:

    APPLE_SDK_VERSION_OVERRIDE=10.13 \

This is fine since it is using Xcode 9.2 which contains only the 10.13 SDK, but that has nothing to do with the deployment target.

[ryandesign (Ryan Schmidt)]

This is still happening, e.g.

• py37-tensorflow1: ​https://build.macports.org/builders/ports-10.12_x86_64-builder/builds/141879/steps/install-port/logs/stdio/text
• py39-tensorflow1: ​https://build.macports.org/builders/ports-10.12_x86_64-builder/builds/141876/steps/install-port/logs/stdio/text

Please fix the port or mark it as unbuildable (known_fail yes) so that the build machine doesn't waste hours of time trying to build this every time the port is modified.

[ryandesign (Ryan Schmidt)]

This is still happening e.g. for py39-tensorflow ​https://build.macports.org/builders/ports-10.12_x86_64-builder/builds/146689

It's really frustrating to waste 15 hours per subport to rediscover this failed build every time any change is made to these ports.

[cjones051073 (Chris Jones)]

I know, I do apologise. I am doing my best to try and address this. I thought I had but clearly not. I suspect I might have to resort to manually renaming the file before installing it, as I am sure it actually would install fine if pip would allow it...

[cjones051073 (Chris Jones)]

​https://github.com/tensorflow/tensorflow/issues/36607

[mascguy (Christopher Nielsen)]

Can we manually kill the build jobs on the 10.12 builder, until a fix is committed? They're simply going to fail anyway...

[cjones051073 (Chris Jones)]

sure, I've been keeping an eye out for them..

[cjones051073 (Chris Jones)]

​https://github.com/macports/macports-ports/commit/f160d7e7c85877c271fb34d45f8541752ae40351

I'm hopeful the above will finally address this...

[cjones051073 (Chris Jones)]

Note, I ideally would set the deployment target bazel uses, but I have failed to get any of the suggestions I can find on how to do this to work...

[mascguy (Christopher Nielsen)]

The fix addressed the issues, and py-tensorflow ports are now building successfully on 10.12. Thanks CJones!

Closing as resolved.