Make new release?

[toaarnio]

Would it be possible to make a new release to PyPI? There are some unreleased changes in the current master that I'm currently having to build manually.

[letmaik]

Absolutely, I'm going to drop Python 2.x support, hope that's fine.

[letmaik]

There is a pre-release version out: 1.14.0a1, please use that in the meantime.

EDIT: I'm going to wait for Python 3.8 and do a new release then.

[toaarnio]

Python 3.8 was released on Oct 14: https://docs.python.org/3/whatsnew/3.8.html

[letmaik]

There are still build failures because not all dependencies have picked up 3.8 yet: https://github.com/letmaik/rawpy/commit/856f4804647c66f4a87f7ccb34424f31d717afc4/checks?check_suite_id=272122321. We'll have to wait a bit longer.

[kmilos]

Btw, would be nice to include lcms2 for Windows in the new release as well, looks like it might happen soon: https://github.com/conda-forge/lcms2-feedstock/pull/3

[hughes]

There are still build failures because not all dependencies have picked up 3.8 yet: https://github.com/letmaik/rawpy/commit/856f4804647c66f4a87f7ccb34424f31d717afc4/checks?check_suite_id=272122321. We'll have to wait a bit longer.

Has this changed in the last few months?

[letmaik]

Unfortunately yes, the following is the dump when trying to install the 32bit build env via conda create --yes --name pyenv_build python=3.8 numpy=1.17 cython jpeg zlib (EDIT: see the message below for more info re 64 bit):

... Solving environment: ...working... Found conflicts! Looking for incompatible packages. This can take several minutes. Press CTRL-C to abort. failed
UnsatisfiableError: The following specifications were found to be incompatible with each other: Package wheel conflicts for: numpy=1.17 -> python[version='>=3.7,<3.8.0a0'] -> pip -> wheel cython -> python[version='>=3.6,<3.7.0a0'] -> pip -> wheel python=3.8 -> pip -> wheel Package wincertstore conflicts for: cython -> setuptools -> wincertstore[version='>=0.2'] numpy=1.17 -> python[version='>=3.7,<3.8.0a0'] -> pip -> setuptools -> wincertstore[version='>=0.2'] python=3.8 -> pip -> setuptools -> wincertstore[version='>=0.2'] Package ca-certificates conflicts for: python=3.8 -> openssl[version='>=1.1.1d,<1.1.2a'] -> ca-certificates cython -> python[version='>=3.6,<3.7.0a0'] -> ca-certificates numpy=1.17 -> python[version='>=3.7,<3.8.0a0'] -> openssl[version='>=1.1.1a,<1.1.2a'] -> ca-certificates Package setuptools conflicts for: python=3.8 -> pip -> setuptools cython -> setuptools numpy=1.17 -> python[version='>=3.7,<3.8.0a0'] -> pip -> setuptools Package certifi conflicts for: cython -> python[version='>=3.6,<3.7.0a0'] -> pip -> requests -> certifi[version='>=2017.4.17'] numpy=1.17 -> python[version='>=3.7,<3.8.0a0'] -> pip -> requests -> certifi[version='>=2016.09|>=2016.9.26|>=2017.4.17'] python=3.8 -> pip -> setuptools -> certifi[version='>=2016.09|>=2016.9.26'] Package python conflicts for: cython -> python[version='>=2.7,<2.8.0a0|>=3.5,<3.6.0a0|>=3.6,<3.7.0a0|>=3.7,<3.8.0a0|>=3.8,<3.9.0a0'] python=3.8 numpy=1.17 -> python[version='>=3.6,<3.7.0a0|>=3.7,<3.8.0a0'] Package msgpack-python conflicts for: cython -> python[version='>=3.6,<3.7.0a0'] -> pip -> cachecontrol -> msgpack-python numpy=1.17 -> python[version='>=3.7,<3.8.0a0'] -> pip -> cachecontrol -> msgpack-python Package pip conflicts for: python=3.8 -> pip numpy=1.17 -> python[version='>=3.7,<3.8.0a0'] -> pip cython -> python[version='>=3.6,<3.7.0a0'] -> pip

To be honest I don't understand where exactly the problem is though. The output is a bit cryptic. EDIT: It's numpy which is missing in a 32bit variant for 3.8 on conda. I'm not sure why that's the case as numpy does have 32bit releases on PyPI.

[letmaik]

It looks like there are missing packages for 32 bit Python 3.8 for the build env (which depends on numpy cython jpeg zlib only) as seen in the dump above (numpy is missing), whereas for 64 bit the build env installs fine but then the test env fails to install (depends on numpy and scikit-image) because a dependency of scikit-image isn't released for 3.8 yet on conda, I'm not sure which one exactly (EDIT: it looks like it's bokeh and by further looking at https://anaconda.org/conda-forge/bokeh/files you can see that the 3.8 package is labelled broken, which is why it hasn't made it's way to the default anaconda channel yet):

Package python conflicts for: scikit-image -> dask[version='>=0.5'] -> bokeh -> requests[version='>=1.2.3'] -> urllib3[version='>=1.21.1,<1.25'] -> ipaddress -> python[version='<=3.3'] scikit-image -> python[version='>=2.7,<2.8.0a0|>=3.5,<3.6.0a0|>=3.6,<3.7.0a0|>=3.7,<3.8.0a0'] numpy -> python[version='>=2.7,<2.8.0a0|>=3.5,<3.6.0a0|>=3.6,<3.7.0a0|>=3.7,<3.8.0a0|>=3.8,<3.9.0a0'] python=3.8 scikit-image -> dask[version='>=0.5'] -> python[version='>=3.5|>=3.6'] Package pip conflicts for: numpy -> python[version='>=3.7,<3.8.0a0'] -> pip python=3.8 -> pip scikit-image -> python[version='>=2.7,<2.8.0a0'] -> pip Package certifi conflicts for: scikit-image -> dask[version='>=0.5'] -> bokeh -> requests[version='>=1.2.3'] -> certifi[version='>=2017.4.17'] scikit-image -> matplotlib[version='>=1.3.1'] -> setuptools -> certifi[version='>=2016.09|>=2016.9.26'] numpy -> python[version='>=3.7,<3.8.0a0'] -> pip -> requests -> certifi[version='>=2016.09|>=2016.9.26|>=2017.4.17'] python=3.8 -> pip -> setuptools -> certifi[version='>=2016.09|>=2016.9.26'] Package msgpack-python conflicts for: numpy -> python[version='>=3.7,<3.8.0a0'] -> pip -> cachecontrol -> msgpack-python scikit-image -> numpy[version='>=1.11.3,<2.0a0'] -> mkl_fft[version='>=1.0.6,<2.0a0'] -> msgpack-python Package wincertstore conflicts for: scikit-image -> matplotlib[version='>=1.3.1'] -> setuptools -> wincertstore[version='>=0.2'] python=3.8 -> pip -> setuptools -> wincertstore[version='>=0.2'] numpy -> python[version='>=3.7,<3.8.0a0'] -> pip -> setuptools -> wincertstore[version='>=0.2'] Package setuptools conflicts for: numpy -> python[version='>=3.7,<3.8.0a0'] -> pip -> setuptools python=3.8 -> pip -> setuptools scikit-image -> matplotlib[version='>=1.3.1'] -> setuptools Package ca-certificates conflicts for: numpy -> python[version='>=3.7,<3.8.0a0'] -> ca-certificates python=3.8 -> openssl[version='>=1.1.1d,<1.1.2a'] -> ca-certificates scikit-image -> python[version='>=2.7,<2.8.0a0'] -> ca-certificates Package wheel conflicts for: numpy -> python[version='>=3.7,<3.8.0a0'] -> pip -> wheel scikit-image -> python[version='>=2.7,<2.8.0a0'] -> pip -> wheel python=3.8 -> pip -> wheel

[letmaik]

I decided to drop 32bit support for Windows as it looks like numpy won't be released in 32bit on conda anymore, and rawpy also hasn't got 32bit releases for Linux and macOS anymore for some time. Now we only need to wait for the Python 3.8 release of bokeh such that the test env can be installed. I opened an issue for that here: https://github.com/conda-forge/bokeh-feedstock/issues/43

[letmaik]

The issue seems to have resolved itself as bokeh was removed as dependency from scikit-image apparently. As soon as #98 is merged, a new rawpy version will be released.

[letmaik]

New release 0.14.0 is being created. Thanks all for the patience!