Depends-on tag

Depends-on tag

[Jeff Johnson]

patchwork itself very recently formalized the usage of a 'Depends-on:' tag for
flagging patchset dependencies:

https://lists.ozlabs.org/pipermail/patchwork/2025-January/007465.html
https://lists.ozlabs.org/pipermail/patchwork/2025-March/007490.html
https://github.com/getpatchwork/patchwork/commit/bc2d1b1dcf89c619e027ae4dfd3036cba30e5583
https://patchwork.ozlabs.org/project/patchwork/list/?series=442332&state=*

Note that this support is only in mainline and not part of any release yet.

Curious if the Linux Kernel community will adopt this approach for flagging
dependencies.

/jeff

Re: Depends-on tag

[Konstantin Ryabitsev]

On Wed, May 28, 2025 at 07:19:55AM -0700, Jeff Johnson wrote:
> patchwork itself very recently formalized the usage of a 'Depends-on:' tag for
> flagging patchset dependencies:
> 
> https://lists.ozlabs.org/pipermail/patchwork/2025-January/007465.html
> https://lists.ozlabs.org/pipermail/patchwork/2025-March/007490.html
> https://github.com/getpatchwork/patchwork/commit/bc2d1b1dcf89c619e027ae4dfd3036cba30e5583
> https://patchwork.ozlabs.org/project/patchwork/list/?series=442332&state=*
> 
> Note that this support is only in mainline and not part of any release yet.
> 
> Curious if the Linux Kernel community will adopt this approach for flagging
> dependencies.

We've been supporting something like this in b4 for a while now, but we use
prerequisite-* trailers (not tags) and always expand that into
prerequisite-patch-id, since that's the only format documented by git itself.

E.g. here is an example of a series using prerequisite-change-id:
https://lore.kernel.org/b4-sent/20250526-sip_svc_upstream-v3-0-6a08a4502de3@altera.com/

Here's another one using prerequisite-message-id:
https://lore.kernel.org/b4-sent/20250514-ipq5018-pcie-v10-0-5b42a8eff7ea@outlook.com/

The documentation is here:
https://b4.docs.kernel.org/en/latest/contributor/prep.html#working-with-series-dependencies

I'm not sure "Depends-on:" as an actual commit tag will be accepted by the
kernel community, because it's information that is only useful during
pre-commit stages. However, if it's passed along as a cover letter/first patch
trailers, we can recognize it as an alternative for "prerequisite-message-id".

-K