> This is far too big for a single patch. It needs to be broken into
> functional chunks that can be reviewed individually. Each driver and
> each device tree change along with it's accompanying code need to be
> done in individual patches. The way it is it can't be reviewed in any
> sane manner.
> -corey
Thanks for your feedback. We are getting a little bit lost here, as our plan was to submit initial
- bindings
- dts for SoC and 1 board
- initial platform init code
Then drivers code avoiding to send many dts updates which might complexify the review. We wanted to send all drivers code to relevant reviewers by tomorrow.
So, what you are asking ( do not worry I am not trying to negotiate, I just want to avoid English misunderstandings as I am French) is to send per driver
- binding
- dts update
- driver code
For each driver through different submission (with each of them containing the 3 associated parts) ?
What shall be the initial one in our case as we are introducing a platform ? An empty dts infrastructure and then we make it grow one step at a time ?
vejmarie
On Wed, Feb 2, 2022 at 7:14 PM Verdun, Jean-Marie <[email protected]> wrote:
>
> > This is far too big for a single patch. It needs to be broken into
> > functional chunks that can be reviewed individually. Each driver and
> > each device tree change along with it's accompanying code need to be
> > done in individual patches. The way it is it can't be reviewed in any
> > sane manner.
>
> > -corey
>
> Thanks for your feedback. We are getting a little bit lost here, as our plan was to submit initial
>
> - bindings
> - dts for SoC and 1 board
> - initial platform init code
>
> Then drivers code avoiding to send many dts updates which might complexify the
> review. We wanted to send all drivers code to relevant reviewers by tomorrow.
>
> So, what you are asking ( do not worry I am not trying to negotiate, I just want
> to avoid English misunderstandings as I am French) is to send per driver
>
> - binding
> - dts update
> - driver code
>
> For each driver through different submission (with each of them containing the
> 3 associated parts) ?
>
> What shall be the initial one in our case as we are introducing a platform ?
> An empty dts infrastructure and then we make it grow one step at a time ?
Ideally, what I prefer to see is a series of patches for all "essential" drivers
and the platform code that includes:
- one patch for each new binding
- one patch for each new driver
- one patch that hooks up arch/arm/mach-hpe/, MAINTAINERS
and any other changes to arch/arm/ other than dts
- one patch that adds the initial .dts and .dtsi files, with all the
devices added that have a valid binding, no need to split this
up any further
This should include everything you need to boot into an initramfs
shell, typically cpu, serial, timer, clk, pinctrl, gpio, irqchip. We will
merge these as a git branch in the soc tree.
In parallel, you can work with subsystem maintainers for the
"non-essential" drivers to review any other driver and binding,
e.g. drm/kms, network, i2c, pci, usb, etc. The patches for
the corresponding .dts additions also go through the soc tree,
but to make things simpler, you can send those in for a later
release.
Arnd
On Wed, Feb 02, 2022 at 06:14:57PM +0000, Verdun, Jean-Marie wrote:
> > This is far too big for a single patch. It needs to be broken into
> > functional chunks that can be reviewed individually. Each driver and
> > each device tree change along with it's accompanying code need to be
> > done in individual patches. The way it is it can't be reviewed in any
> > sane manner.
>
> > -corey
>
> Thanks for your feedback. We are getting a little bit lost here, as our plan was to submit initial
>
> - bindings
> - dts for SoC and 1 board
> - initial platform init code
>
> Then drivers code avoiding to send many dts updates which might complexify the review. We wanted to send all drivers code to relevant reviewers by tomorrow.
>
> So, what you are asking ( do not worry I am not trying to negotiate, I just want to avoid English misunderstandings as I am French) is to send per driver
>
> - binding
> - dts update
> - driver code
>
> For each driver through different submission (with each of them containing the 3 associated parts) ?
Arnd gave an excellent explaination for this.
To be clear, you need to split out changes to individual subsystems and
submit those to the maintainers for that subsystem and not send them to
everyone. That way you reduce sending emails to people who don't need
to see them.
Once you have a set of patches for a subsystem, you can submit them as one
set. That is generally preferred. The "git send-email" or "git
format-patch" tools are generally what we use, they let you compose a
header message where you can give an overall explaination, then it sends
the individual changes as followup messages to the header message.
-corey
>
> What shall be the initial one in our case as we are introducing a platform ? An empty dts infrastructure and then we make it grow one step at a time ?
>
> vejmarie
>
>
>
>
>
> _______________________________________________
> Openipmi-developer mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/openipmi-developer