search

conda-forge / osmnx-feedstock

A conda-smithy repository for osmnx.
BSD 3-Clause "New" or "Revised" License
8 stars 7 forks source link

re-render feedstock, bump build number #40

Closed gboeing closed 5 years ago

gboeing commented 5 years ago

Checklist

conda-forge-linter commented 5 years ago

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

gboeing commented 5 years ago

same error as was happening in https://github.com/conda-forge/geopandas-feedstock/pull/49

ocefpaf commented 5 years ago

This one should pass as soon as we merge https://github.com/conda-forge/geopandas-feedstock/pull/49

gboeing commented 5 years ago

Looks like same error is still happening?

ocefpaf commented 5 years ago

Looks like same error is still happening?

The new geopandas is not up yet. I'll restart the CI here when it is available in the channel. (The new Anaconda.org, that uses cloudflare, has a delay to get the packages available. You can check with conda search geopandas | grep 0.4.1.)

ocefpaf commented 5 years ago

OK. Now we are pulling geopandas 0.4.1-py_0 and the problem persists :-(

Investigating...

ocefpaf commented 5 years ago

I can confirm that, even with https://github.com/conda-forge/geopandas-feedstock/pull/49, the problem remains. When doing:

conda create --name TEST python=3.7 geopandas

We still get libgdal 2.3.3 from defaults :unamused:

I'm going to try to "force" strict channel here just as an experiment. Hold on...

ocefpaf commented 5 years ago

While this is passing one will only be able to install osmnx when using the strict channel option.

We are between a rock and a hard place here and the GIS stack is pretty much broken without the strict channel option.

PS: a re-rendering will remove that change. I'll discuss with the @conda-forge/core to make that default in conda-smithy but maybe @msarahan will beat us by making it default in conda 4.7 :wink:

gboeing commented 5 years ago

While this is passing one will only be able to install osmnx when using the strict channel option.

Quick question about strict channel priority. If I run:

conda create -n TEST -c conda-forge --strict-channel-priority geopandas

with conda 4.6.7 on windows 10, I get:

Collecting package metadata: done
Solving environment: failed

UnsatisfiableError: The following specifications were found to be in conflict:
  - geopandas
Use "conda search <package> --info" to see the dependencies for each package.

I tested a few other packages, and the same thing happens if I replace geopandas with statsmodels, cartopy, or seaborn. Other packages however do solve successfully (bokeh or pillow, for example). Are these UnsatisfiableErrors expected? If so, how do I install these packages on windows with strict channel priority?

ocefpaf commented 5 years ago

I tested a few other packages, and the same thing happens if I replace geopandas with statsmodels, cartopy, or seaborn. Other packages however do solve successfully (bokeh or pillow, for example). Are these UnsatisfiableErrors expected? If so, how do I install these packages on windows with strict channel priority?

What platform is that? Windows or macOS? I can only test on Linux and all the packages you mentioned worked for me. I do know that on Windows we cannot use strict b/c conda-forge lacks a package for scipy. Not sure about macOS.

gboeing commented 5 years ago

conda 4.6.7 on Windows 10

gboeing commented 5 years ago

So if the GIS stack is broken without the strict option, and if strict doesn't work on Windows, how do we install on Windows at this point?

ocefpaf commented 5 years ago

So if the GIS stack is broken without the strict option, and if strict doesn't work on Windows, how do we install on Windows at this point?

Windows is a corner case where conda-forge relies a lot on defaults and therefore the exception to the strict channel. We should make that more explicit in the docs. macOS and Linux should prefer strict and Windows, until we package scipy, should not.

gboeing commented 5 years ago

Got it. Thanks @ocefpaf

CJ-Wright commented 5 years ago

@ocefpaf shouldn't strict have noticed that we don't have a windows scipy and properly pulled deps from defaults?

ocefpaf commented 5 years ago

@ocefpaf shouldn't strict have noticed that we don't have a windows scipy and properly pulled deps from defaults?

In theory yes. I test that and it did not work. Probably more unsatisfied deps and/or a bug in strict.

ocefpaf commented 5 years ago

Got it. Thanks @ocefpaf

I'm sorry that the state of things are half-broken/half-quasi-working but we are working on using the strict channel across conda-forge, see https://github.com/conda-forge/conda-forge-ci-setup-feedstock/pull/50, and hopefully fix the SciPy dependency on Windows.

gboeing commented 5 years ago

Thanks! In the meantime, I'll tell linux/mac folks to use strict, and windows folks to not.

ocefpaf commented 5 years ago

Thanks! In the meantime, I'll tell linux/mac folks to use strict, and windows folks to not.

Thanks!

gboeing commented 5 years ago

uh-oh, this failed upon merge: https://circleci.com/gh/conda-forge/osmnx-feedstock/92

gboeing commented 5 years ago

I don't think I have permissions to re-start the CI workflow.

ocefpaf commented 5 years ago

I don't think I have permissions to re-start the CI workflow.

Just did it, but you should have as a maintainer :-/