Replace cargo-generate with kickstart

[eupn]

Closes #36.

An example of prompts with defaults:

❯ kickstart https://github.com/knurling-rs/app-template
Author's name? [default: You <[email protected]>]: 
How to you name your project's crate? [default: my-embedded-app, validation: ^([a-zA-Z][a-zA-Z0-9_-]+)$]: 
What's the name of the chip (see `probe-run --list-chips` output)? [default: nRF52840_xxAA]: 
What's the target's name?: 
  1. thumbv6m-none-eabi (Cortex-M0 and Cortex-M0+)
  2. thumbv7m-none-eabi (Cortex-M3)
  3. thumbv7em-none-eabi (Cortex-M4 and Cortex-M7 | no FPU)
  4. thumbv7em-none-eabihf (Cortex-M4F and Cortex-M7F | with FPU)
  > Choose from 1..4 [default: 1]: 
Do you want to use an embedded-hal?: 
  1. yes
  2. no
  > Choose from 1..2 [default: 1]: 
embedded-hal crate name? [default: nrf52840-hal]: 
embedded-hal crate version? [default: 0.11.0]: 

Everything done, ready to go!

Tested by compiling the generated project with defaults.

[Sympatron]

I think this would make it a bit easier to start.
Can you use a URL with kickstart for the template? With cargo generate you could specify --git https://github.com/knurling-rs/app-template so you don't have to clone the template first.

I think removing setup step 3 and 5 is fine, but the rest of the steps contain information that is still valuable. I would just change the order, because the user has to lookup chip name and HAL name and version before starting kickstart.

[eupn]

@Sympatron

Hello and thank you for the feedback!

Can you use a URL with kickstart for the template?

Yes, I've updated the README with a very similar command to use with kickstart:

$ kickstart https://github.com/knurling-rs/app-template

I would just change the order, because the user has to lookup chip name and HAL name and version before starting kickstart.

Yeah, it makes sense to first introduce the probe-run's capabilities in terms of supported chips before suggesting kickstart commands.

[japaric]

Thanks for the PR, @eupn. The UI looks good.

Something I found very surprising is that kickstart $URL copies the template contents into the current working directory instead of creating a new folder. Is there a way to make it generate a new folder like cargo-generate does? If not, let's please document that behavior in the README.

I noticed that kickstart shells out to git to fetch a git repository. We should document in the README that having git installed is a pre-requisite. Not all our users have git installed (e.g. on Windows) and we have already applied work around to make e.g. defmt work without the git tool so we should give users a "heads up" here.

I would prefer to keep the 'how to do this manually' info in the README as some people may prefer to download a zip version of this repository rather than clone it. We can divide the README in two sections: kickstart and "manual setup".

(Just thinking out loud: the other thing I noticed is that if you answer 'no' to 'do you want to use an embedded-hal?' then the generated project will fail to build with "memory.x file not found". I guess if one takes that path the template should also generate the memory.x file -- though answering how much Flash / RAM can require some digging and may be daunting for beginners. (Then again some HAL crates don't provide a memory.x either))

[eupn]

@japaric

Something I found very surprising is that kickstart $URL copies the template contents into the current working directory instead of creating a new folder. Is there a way to make it generate a new folder like cargo-generate does? If not, let's please document that behavior in the README.

You can specify the output directory via --output-dir <dir> flag. I'll include this in README for it to be less surprising.

I noticed that kickstart shells out to git to fetch a git repository. We should document in the README that having git installed is a pre-requisite. Not all our users have git installed (e.g. on Windows) and we have already applied work around to make e.g. defmt work without the git tool so we should give users a "heads up" here.

Sure, I'll document this as well.

I would prefer to keep the 'how to do this manually' info in the README as some people may prefer to download a zip version of this repository rather than clone it. We can divide the README in two sections: kickstart and "manual setup".

Yes, let's keep the manual steps and make two sections in README.

(Just thinking out loud: the other thing I noticed is that if you answer 'no' to 'do you want to use an embedded-hal?' then the generated project will fail to build with "memory.x file not found". I guess if one takes that path the template should also generate the memory.x file -- though answering how much Flash / RAM can require some digging and may be daunting for beginners. (Then again some HAL crates don't provide a memory.x either))

Not sure what would be the preferred approach here if the user doesn't want to use HAL crate. Probably we'll just let the user write the memory.x file themselves by consulting their HAL's docs.

[jonas-schievink]

We support thumbv8 too

[Urhengulas]

Hi @eupn, Can you please rebase your PR and resolve the merge conflicts?

[eupn]

@Urhengulas sure!