Vendor provided code might have a different license, different copyrights, or 
maybe open-but-proprietary and could not come into our Apache 2 repositories.

Nor would vendor code likely follow the coding standard and, hence, not pass CI 
checks.

I am not deeply involved in this decision, but I don't see most vendor provided 
HALs integrating into the NuttX ASF repositories without a massive effort.  
That is why the github NuttX repository was created:  to work around licensing 
and ownership issues of the AFS repositories.

In short, I hear only two options in this discussion:


  *
Expend the effort to make the HAL fully integrated into the ASF repository and 
CI, or
  *
Use an external repository that is not bound to follow ASF / NuttX standards.

I suppose we could also add the one I never liked but we did a few times in the 
past:


  *
Dynamically download vendor HALs at build time.  One of the AVRs used to do 
that.  ESP32's used to do that too.

________________________________
From: Alan C. Assis <[email protected]>
Sent: Thursday, September 24, 2026 6:28 AM
To: [email protected] <[email protected]>
Subject: Re: Early feedback on RZ/V2H port and RZ FSP HAL dependency

Greg, how can we enforce the vendors to follow a standard? If it is a HAL,
it is specific to each vendor.

Forcing them to follow a standard is like forcing them to write native code
for NuttX.

Perhaps following CMSIS is an option, but it is ARM-only, there is not a
universal HAL.

BR,

Alan

On Thu, Sep 24, 2026 at 9:39 AM Gregory Nutt <[email protected]> wrote:

> This repository can follow some common rules so
> different HALs wouldn't be completely different and we would have some
> control over it.
>
> That is what we did in the past to handle devices (mostly wireless
> devices) that required a firmware download each time they booted.  We kept
> the open-but-proprietary firmware in a separate repository known to the
> build system
>
> (Since that repository is not generally available I imagine that those old
> parts are no longer usable).  The disconnection of the OS from the
> supported repository was ultimately fatal.
>

Reply via email to