-
Notifications
You must be signed in to change notification settings - Fork 47
Strategy for multiple binary options per platform? #155
Comments
Mmmm, yes, this gets tricky. Everything is going to have to go through Pkg eventually, so we're going to need to figure out what the best strategy for this is long-term. I'm going to ping @StefanKarpinski and @KristofferC to get their opinion on what they think the best way to get these kinds of "runtime adaptive dependencies" are. Short-term, you can do whatever runtime inspection you need within |
@staticfloat I like the multiple One question, if I prompt the user for options during the build process, it would make CI builds hang until timeout. Is there a user input with timeout method available? i.e.
|
Not that I am aware. You could override it with an environment variable. That being said, I'm not sure you can interact with the user during |
Ah of course.. So the selection would have to happen at |
I have a library that has different build processes depending on whether CUDA, cuDNN, tensor cores, OpenMP etc. are available on the build system, AND whether you want to use them..
So, I'm building a BinaryProvider-based wrapper for this library and want to give the user the option of 2-4 binaries per platform. Something like:
The simplest thing would be to give the user the option and try to install and test, with some generic message telling the user to make sure those dependencies are present.
The safer thing is to give the option, test for presence of dependencies, then install and test.
Are there any good examples of how to do this with BinaryProvider?
@staticfloat ?
The text was updated successfully, but these errors were encountered: