Rajath c S
Rajath c S

Reputation: 1

Building packages to yocto image (image) without using recipes

I am a newbie to yocto. Have worked on its basics for few weeks and now I have the task of installing few packages on to the yocto image (poky) and I'm completely stuck and don't know what to do.

The procedure I am following is : I have cloned the packages from github on to my host machine and put inside a directory. Now in poky I have deb , apt-get working. But how do I install these cloned packages . I do not want to follow the recipe /layers method. The packages i want to put is : numpy, tensor flow, pyserial, h5py.

Upvotes: 0

Views: 2027

Answers (1)

LetoThe2nd
LetoThe2nd

Reputation: 1326

Technically, you can generate a SDK or eSDK for the image that you use, and with that manually build and deploy to the target. Yet while his is technically possible, it has a lot of downsides:

  • you are working around package management
  • what you do is totally unreproductible
  • you will run into lots and lots of problems to fix manually that using the correct mechanism would spare you from, as it already tackles those properly. Especially when compilation is necessary, there are a lot of pitfalls.

You can find the respective documentation here: https://www.yoctoproject.org/docs/2.4.2/mega-manual/mega-manual.html#sdk-using-the-standard-sdk and https://www.yoctoproject.org/docs/2.4.2/mega-manual/mega-manual.html#sdk-extensible

So in a nutshell: the real point of Yocto/OpenEmbedded is exactly the recipes/metadata mechanisms that you want to not use. If that is a hard requirement for whatever reason (not just "hey I don't like writing recipes"), then the Yocto Project technology is not a good fit for what you need. If you actually want debian-style package management, why not just use debian? Nothing wrong with that, if it fits your usecase.

Upvotes: 1

Related Questions