gpanders / meta-scipy

OpenEmbedded layer for scipy (ARCHIVED)
https://github.com/tuxable-ltd/meta-scipy
MIT License
18 stars 23 forks source link

Add a layer.conf #3

Closed mcampbellizo closed 4 years ago

mcampbellizo commented 4 years ago

This PR adds a layer.conf so this can be integrated with bitbake in a full Yocto build. This was borrowed from the @HerrMuellerluedenscheid fork of this repo. Looks like the three of us are all trying to solve the same problem of getting scipy into Yocto. Any thoughts on how we can best centralize and share our efforts? I'm happy to discuss in the PR how you or @HerrMuellerluedenscheid feel about this.

Slightly related to this PR, we prefer @HerrMuellerluedenscheid's repo naming of meta-scipy which follows the standard layer naming convention. Do you have any thoughts about switching to the more standard naming convention?

gpanders commented 4 years ago

Any thoughts on how we can best centralize and share our efforts? I'm happy to discuss in the PR how you or @HerrMuellerluedenscheid feel about this.

You're of course welcome to contribute here as much as you like. Once the recipe is "stable" enough, I'd like to submit it for integration into the mainline meta-openembedded layer.

Slightly related to this PR, we prefer @HerrMuellerluedenscheid's repo naming of meta-scipy which follows the standard layer naming convention. Do you have any thoughts about switching to the more standard naming convention?

I actually intentionally did not choose meta-scipy for the name of this repo because I did not intend for it to be used as a separate layer in and of itself, but rather to simply be copied into an existing layer. My long term intention was to host the recipe here for others to use and eventually submit it upstream (as I mentioned above).

That being said, it seems perfectly reasonable to me that people would want to use this as a separate layer, so adding a layer.conf file seems like a good idea, but I'd like to keep the name as oe-scipy to emphasize that it's not intended to be its own layer.

HerrMuellerluedenscheid commented 4 years ago

If I may add: When I was forking the repo, renaming to meta-scipy and adding the layer.conf I simply could not make it work in another way mostly due to lack of experience. After you added the (noob-)installation details in the README.md I actually do not use my repo fork anymore. I'll archive it in a minute to avoid confusion with the main branch. Having this recipe in it's own layer appears nice to me but I also understand your point @gpanders that this is rather an intermediate step before submitting this to meta-oe, which I and many others will certainly will highly appreciate!!!

HerrMuellerluedenscheid commented 4 years ago

Regarding naming: Convention is to prepend meta to a layer name which you certainly are aware of and for good reason deviate from. However, there is one more benefit sticking to that convention: People will find this layer/recipe much faster as it will be much more obvious and visible. How about I keep it archived as "meta-scipy" pointing users to the main line repo? checkout https://github.com/HerrMuellerluedenscheid/meta-scipy

gpanders commented 4 years ago

After sleeping on it, I think I've had a change of heart. I anticipate that getting this recipe merged into the upstream meta-openembedded layer will not be easy or straightforward (or maybe not possible at all) given how many other recipes need to be modified to get scipy to cross-compile.

Given that, I think the best path forward for now is to offer this as a separate layer. We can resolve the other issues if/when the time comes that this gets enough attention and is tested enough to warrant trying to get it merged upstream.

gpanders commented 4 years ago

Closed in 5fa8b4a. Thanks!

mcampbellizo commented 4 years ago

I agree with the ultimate goal of getting this in meta-openembedded, but also that it may take some time to get there. In the mean time, I love the idea of fleshing out this layer here so we can continue to mature this and also have a place for others to use/contribute that need scipy in Yocto/OE.

This was a great discussion and it was a pleasure working with you both!