Open bsipocz opened 7 years ago
I saw that conda-channel-astropy
moved to conda-forge, and I'm fine with doing the same thing for the packages I created here. What's the process to create a new package on conda-forge ?
I've never done it, but my understanding is first to create a PR with the recipe at https://github.com/conda-forge/staged-recipes/, then when it's merged it becomes a "feedstock", a separate repo where you (the people listed in the recipe) become maintainer, and all the version dependent updates (mostly only version number and hash) happen there.
@bsipocz -- I can plan to convert this next week. It is not entirely clear to me it makes to maintain it as a separate channel anymore.
@saimn -- create a recipe, open a PR at the repo @bsipocz mentioned. if you have conda-build 3.0.8 or higher then you can use conda skeleton to generate a recipe that just needs light editing to get it into the format that conda-forge wants. Usage is: conda skeleton pypi your_package_name
Thanks for the hints ! I will have a look.
@dpshelio @Cadair - Shall we archive this repo? It looks like no update has happened to the packages for at least 4 years, and the standard these days anyway is to use conda-forge
for everything.
(I still like the idea of having a meta channel for the openastronomy packages, but if we don't have the capacity to maintain it (and have no better reason that "would be nice", then it's probably better to archive than keep it open as is.)
I am pro archiving this, it's very dead :grinning:
Should we do any of
I think the only one of those that is essential is to add a note to the README directing folks to the conda-forge channel. Once the repo is archived the issues/pulll requests will be locked.
It seems that the infrastructure here is starting to break horribly (see e.g. #35), and it's very much diverged from what we had in astropy/conda-channel-astropy anyway. Do we have a vision for future directions here?
Should we switch to the same model as we do in astropy, namely build the packages on conda-forge and then copy the latest versions over with a travis cron job? Does it worth maintaining a totally separate infrastructure here?
cc @Cadair @mwcraig @dpshelio @eteq @astrofrog (I'm sorry for the excessive pinging, but I'm actually not sure who is supposed to maintain this.)