Re: [yocto-patches] [meta-rockchip] Backport patches to Whinlatter

Martin Herren <[email protected]> Mon, 20 Apr 2026 17:27:55 +0000

Newsgroups	org.yoctoproject.lists.yocto-patches
Message-ID	<D_T0vfLiZ7iT-Zdru_yLBe3ZScLI_weV-fDDAFnDD4aJybbieGUfMXTEA-X5vakiZbsQeTAZ3DjRdyJRCZjM2FxdcR1WwxO4rk1cUUCTg7w=@on-the-web.ch>

Hej,

> Hi Martin, Trevor,
> 

> On 4/20/26 12:14 PM, Martin Herren wrote:
> > Dear maintainers,
> >
> > May I suggest the following 2 patches from meta-rockchip master to be back ported to the Whinlatter branch ?
> >
> > fitimage rework: fe7cfa748bb132cfbe322c1c6d586f09557dbeb2
> >
> 

> This one is tricky. We shouldn't break existing users but this is a
> patch that isn't backward compatible as we would require users to set
> the RK_KERNEL_FITIMAGE variable to match the modifications they made to
> KERNEL_IMAGETYPE/KERNEL_CLASSES in their own machine configuration file
> based on the SoC machine include file. Can you tell us more about what
> is broken before this is backported? Maybe there's another way to
> implement a fix without breaking backward compatibility.

Doing a `bitbake core-image-base` for `MACHINE = "rock-5b"` on current whinlatter (including all other layers also on whinlatter) without this patch leads to:

```
❯ bitbake core-image-base
Loading cache: 100% |#######################################################################################################################################################| Time: 0:00:00
Loaded 0 entries from dependency cache.
ERROR: ParseError at /home/martin/dev/yocto/rockchip/layers/openembedded-core/meta/classes-recipe/kernel.bbclass:188: Could not inherit file classes/kernel-fitimage.bbclass| ETA:  0:00:17
ERROR: Parsing halted due to errors, see error messages above

Summary: There were 2 ERROR messages, returning a non-zero exit code.
```

Applying this patch goes further:

```
ERROR: rockchip-rkbin-tf-a-git-r0 do_unpack: Recipes that set S = "${WORKDIR}/git" or S = "${UNPACKDIR}/git" should remove that assignment, as S set by bitbake.conf in oe-core now works.
ERROR: Logfile of failure stored in: /home/martin/dev/yocto/rockchip/layers/build/tmp/work/rock_5b-oe-linux/rockchip-rkbin-tf-a/git/temp/log.do_unpack.34439
ERROR: Task (/home/martin/dev/yocto/rockchip/layers/meta-rockchip/recipes-bsp/rkbin/rockchip-rkbin-tf-a_git.bb:do_unpack) failed with exit code '1'
ERROR: rockchip-rkbin-ddr-git-r0 do_unpack: Recipes that set S = "${WORKDIR}/git" or S = "${UNPACKDIR}/git" should remove that assignment, as S set by bitbake.conf in oe-core now works.
ERROR: Logfile of failure stored in: /home/martin/dev/yocto/rockchip/layers/build/tmp/work/rock_5b-oe-linux/rockchip-rkbin-ddr/git/temp/log.do_unpack.34438
ERROR: Task (/home/martin/dev/yocto/rockchip/layers/meta-rockchip/recipes-bsp/rkbin/rockchip-rkbin-ddr_git.bb:do_unpack) failed with exit code '1'
```

Thus the 2nd suggested backport.

> > bsp: rkbin: fix build since S = "${WORKDIR}/git" is implied by OE-Core: e59f6fcb66091ea5efe33f3f3d8be09b1f65bbfb
> >
> 

> git tag --contains 46480a5e6674
> (the commit listed in the commit log) returns the first whinlatter tag,
> so should be fine to backport.
> 

> > I understand Whinlatter to soon enter EOL, it would make sense to have it enter EOL in a buildable state.
> >
> 

> Yes, but not by potentially breaking existing users, so special care
> needs to be taken. This is not a worry when we're working on the master.

Yes, didn't think about that it could break other custom board bsp layers based upon this one.

/Martin

signature.asc (application/pgp-signature, 343 B)