build error beginning (07/01/2019)

[jGaboardi]

Current build errors apparently stemming from geopandas/conda-forge on Python 3.7. This same error also seems to be affecting meta pysal, libpysal, and segregation.

@renanxcortes @knaaptime @weikang9009

[sjsrey]

Not sure what exactly the issue is, but I think it is related to the travis specification as I tried a number of local installs using conda environemnts w/wo pip and python3.7 and geopandas and can't reproduce the error. In other words, I can build these environments locally just fine.

[knaaptime]

ran into this last week in the sprint with esau. I think it has do do with some dependency conflicts that arise from channel_priority=strict. i.e. i think there are some packages that have yet to be released for 3.7 on conda-forge, and since channel priority is strict, conda refuses to get the appropriate versions from defaults

[weikang9009]

agree with @knaaptime . removing channel_priority=strict sounds promising since the error message points to:

[jGaboardi]

@weikang9009 However, if channel_priority=strict is removed then the install of geopandas-related dependencies gets funky (unless this has been addressed very recently). I think we should let this one ride for the time being and hope a solution crops up from the conda-forge folks. Thoughts?

@ocefpaf

[knaaptime]

that was my inclination

[ocefpaf]

@weikang9009 However, if channel_priority=strict is removed then the install of geopandas-related dependencies gets funky (unless this has been addressed very recently). I think we should let this one ride for the time being and hope a solution crops up from the conda-forge folks. Thoughts?

conda-forge always recommend using strict when adding a third party channel. If you remove it things may work but there is no guarantee that the packages are compatible with each other (compilers, dependency version, etc). I'll look into the original issue to see what may be going on there.

[ocefpaf]

Filed an issue at conda/conda#8858

Note that when conda-forge finished the readline migration this CI will be working again but the bug in conda still exists.

[knaaptime]

thanks for your help @ocefpaf!

[ocefpaf]

BTW, I'm trying to work on the readline migration in conda-forge and hopefully this will be fixed by tomorrow.

[ocefpaf]

Looks like all the packages needed for this project were migrated to the new readline. When you restart the CIs let me know how it goes.

[jGaboardi]

@ocefpaf my personal branch of spaghetti is now passing and the main pysal/pysal is also passing. Other submodules using geopandas are currently running, but it looks like we are all good to go. Thanks again for the support!

[jGaboardi]

All looks in order.