summaryrefslogtreecommitdiffstats
path: root/conf
diff options
context:
space:
mode:
authorLuciano Dittgen <luciano.dittgen@ossystems.com.br>2026-08-06 15:05:47 -0300
committergithub-actions[bot] <github-actions[bot]@users.noreply.github.com>2026-08-09 00:38:12 +0000
commit1763777455a08e31db1309fb7944ae03589b90ee (patch)
treee4494fbbdfd3f399f291d5dce2d6b68f48eb0d80 /conf
parent346808a521c4510588466bfcc4f46f5bb3fba6c3 (diff)
downloadmeta-freescale-1763777455a08e31db1309fb7944ae03589b90ee.tar.gz
imx-base: Move the i.MX SDK host tools to TOOLCHAIN_HOST_TASK
A bbappend to nativesdk-packagegroup-sdk-host added nativesdk-imx-usb-loader, nativesdk-mxsldr, nativesdk-u-boot-mkimage and nativesdk-wayland to RDEPENDS unconditionally, so an SDK built for any machine in a build with meta-freescale in BBLAYERS picked up i.MX flashing tools. Measured on qemuarm64 with bitbake -e: RDEPENDS gained exactly those four. The obvious fix -- scoping that RDEPENDS to a machine override -- would be worse than the leak. That packagegroup has a fixed PACKAGE_ARCH of ${SDK_ARCH}-nativesdk, so making its contents depend on MACHINE would emit different package content under one package architecture, and two SDKs built from the same TMPDIR would overwrite each other in the feed. oe-core's own recipe is careful here: it varies only by SDK_ARCH and DISTRO_FEATURES, never by MACHINE. Drop the bbappend and add the tools to TOOLCHAIN_HOST_TASK in imx-base.inc instead. populate_sdk_base consumes that variable in the SDK recipe, which is already machine-specific, so there is no package-architecture hazard. imx-base.inc is required only by i.MX machine configurations -- 27 of them, no QorIQ -- and the append is additionally scoped to imx-generic-bsp to match the IMAGE_CLASSES line above it. Measured with bitbake -e, before vs after: qemuarm64 TOOLCHAIN_HOST_TASK gains none of the four tools imx6qdlsabresd TOOLCHAIN_HOST_TASK gains exactly the four tools; nothing else in the expansion moves Note for QorIQ: those machines previously received these tools through the unconditional bbappend and no longer do. Three of the four are i.MX-specific, so this is intended, but it is a behaviour change for machines this layer does own. bitbake -p: 1166 recipes, 0 errors. No SDK was built, so the effect on a populated SDK is inferred from TOOLCHAIN_HOST_TASK, not measured. Signed-off-by: Luciano Dittgen <luciano.dittgen@ossystems.com.br> (cherry picked from commit c42a38609d6bebf15643e3a04acfe42bd8aac117)
Diffstat (limited to 'conf')
-rw-r--r--conf/machine/include/imx-base.inc12
1 files changed, 12 insertions, 0 deletions
diff --git a/conf/machine/include/imx-base.inc b/conf/machine/include/imx-base.inc
index 21cc4ea52..8030d69f9 100644
--- a/conf/machine/include/imx-base.inc
+++ b/conf/machine/include/imx-base.inc
@@ -698,6 +698,18 @@ HOSTTOOLS_NONFATAL:append:mx8-nxp-bsp = " sha384sum"
698# Add task to generate the mfgtool bundle for the image. 698# Add task to generate the mfgtool bundle for the image.
699IMAGE_CLASSES:append:imx-generic-bsp = " image_populate_mfgtool" 699IMAGE_CLASSES:append:imx-generic-bsp = " image_populate_mfgtool"
700 700
701# i.MX host tools for the SDK. Deliberately not a bbappend to
702# nativesdk-packagegroup-sdk-host: that packagegroup has a fixed PACKAGE_ARCH
703# (${SDK_ARCH}-nativesdk), so varying its RDEPENDS by MACHINE would emit
704# different content under one package architecture. TOOLCHAIN_HOST_TASK is
705# consumed by the SDK recipe, which is already machine-specific.
706TOOLCHAIN_HOST_TASK:append:imx-generic-bsp = " \
707 nativesdk-imx-usb-loader \
708 nativesdk-mxsldr \
709 nativesdk-u-boot-mkimage \
710 ${@bb.utils.contains('DISTRO_FEATURES', 'wayland', 'nativesdk-wayland', '', d)} \
711"
712
701SECO_FIRMWARE_NAME:mx943-generic-bsp ?= "mx943${IMX_SOC_REV_LOWER}-ahab-container.img" 713SECO_FIRMWARE_NAME:mx943-generic-bsp ?= "mx943${IMX_SOC_REV_LOWER}-ahab-container.img"
702 714
703# Allow meta-imx to add NIP devices information until upstreamed. 715# Allow meta-imx to add NIP devices information until upstreamed.