Attempt at providing a noblas. #2
Conversation
|
Hi! This is the friendly automated conda-forge-linting service. I wanted to let you know that I linted all conda-recipes in your PR ( Here's what I've got... For recipe:
|
|
It is. Let me try a PR that should help. |
|
So, I'm thinking about this and honestly I would be ok with a |
|
Yeah, I think we will need a patch. Opened a bug report ( numpy/numpy#7652 ) upstream. Let's see if they have any advice. |
| @@ -13,6 +14,9 @@ source: | |||
|
|
|||
| build: | |||
| number: 100 | |||
There was a problem hiding this comment.
Let's go ahead and bump this to 101. Going forward I would like us to reserve 100-199 for noblas and 200-299 for openblas. For this iteration it will have to be 101-109 for noblas. Basically build number is playing a big role in terms of what gets installed. So, I'd like to keep it higher. Again pinning should give you what you want here.
There was a problem hiding this comment.
ping @mcg1969 - this makes me cringe as a method of determining which variant is installed. Please try to find a better way.
There was a problem hiding this comment.
It makes us cringe too, but unfortunately we had to get something ASAP. So this is where @pelson and I are with it. Features with exclusions or selectors is what we really want to see. Being susceptible to every change in version or build number in defaults is really hurting us here.
There was a problem hiding this comment.
How ASAP is ASAP? We're doing channel priority handling for conda 4.1.
There was a problem hiding this comment.
Would have to ask @pelson, but I go the impression that it was highly desirable to have this out by today.
There was a problem hiding this comment.
I mean we can limp along with this for a little while. We tried to pick some that worked now, but should be pretty stable for awhile (even if it is hideous). That way we have some time to work on something better for conda and/or conda-build.
There was a problem hiding this comment.
Just to clarify what was needed ASAP was NumPy and by extension SciPy. We can continue to discuss conda and conda-build changes. What we have done here is not pretty, but it let's these packages come into existence in conda-forge without putting undue pressure on anyone else. We fully expect to change this in the long term.
|
@pelson closing this b/c the rebase alone would be a nightmare. We should try to do this again at some point though. |
Unfortunately numpy is still using accelerate locally, and I suspect it is also the case on the CI.
Useful script: https://gist.github.com/alimuldal/eb0f4eea8af331b2a890