linux/meta-freescale.git/recipes-graphics, branch master

imx-gpu-sdk: Quote DEPENDS_BACKEND and WINDOW_SYSTEM on the assignment line

2026-09-04T18:46:13+00:00

A line continuation before the opening quote hides the value from every reader and tool that looks at the assignment line. Only the Python source inside ${@...} is reflowed, so both values are unchanged. Tested with bitbake imx-gpu-sdk -C configure on imx8mm-lpddr4-evk (fslc-wayland), where the wayland branch of both expressions is taken; the imx-gpu-sdk buildhistory is byte-identical. Signed-off-by: Luciano Dittgen

imx-gpu-viv: Quote PACKAGES_OPENVX on the assignment line

2026-09-04T16:49:53+00:00

A line continuation before the opening quote hides the value from every reader and tool that looks at the assignment line, and a stray space after the backslash breaks the parse silently. Only the Python source inside ${@...} is reflowed, so the value is unchanged. Tested with bitbake imx-gpu-viv -C unpack -c package on imx8mp-lpddr4-evk, where PROVIDES_OPENVX is set; buildhistory is byte-identical. Signed-off-by: Luciano Dittgen

piglit: Suppress noncoreoverride on the upstream cl test patches

2026-09-04T12:52:28+00:00

Both patches fix piglit's OpenCL tests generally, not on i.MX: 0001 corrects an install-layout include path in a cl test, 0002 adds the missing CL 2.0 cl_mem_flags mutexes. Neither touches a compile path, so scoping them to imx-generic-bsp would leave the tests broken everywhere else rather than fix a build. Both carry an upstream MR link, so the divergence ends when they merge. Signed-off-by: Luciano Dittgen

Merge pull request #2664 from lucianogdittgen/oelint-var-licenseremotefile

2026-09-03T18:53:06+00:00

oelint: Resolve oelint.var.licenseremotefile findings (10 -> 0)

mesa-etnaviv-env: Checksum the license header in the shipped script

2026-09-03T14:52:26+00:00

SRC_URI ships only a conf file and a shell script from this layer, with no license header to checksum, so the recipe checksummed oe-core's generic MIT text, which never changes and so verified nothing. Declare MIT with an SPDX identifier in both shipped files and checksum the line in the one do_install actually ships under use-mainline-bsp+sysvinit. Validated on MACHINE=imx8mp-ddr4-evk with IMX_DEFAULT_BSP forced to mainline, the only override that installs mesa-etnaviv.sh. Intended output change: the shipped script grows by the SPDX line, so PKGSIZE goes 63 -> 94. FILELIST is unchanged. Signed-off-by: Luciano Dittgen

libxdg-shell: Drop the variables from the license path

2026-09-03T14:31:36+00:00

LIC_FILES_CHKSUM addressed License.md through ${UNPACKDIR}/${BP}, which a standalone analyzer cannot resolve. A file:// path is already relative to S, and S sits three levels below the source root, so write it that way. The file and the md5 do not change. Validated on MACHINE=imx8mp-ddr4-evk: do_populate_lic re-ran and passed, so the path still resolves; the buildhistory packages tree is identical. Signed-off-by: Luciano Dittgen

devil: Checksum the license the source ships

2026-09-03T14:31:36+00:00

LIC_FILES_CHKSUM pointed at oe-core's generic LGPL-2.1-only text, which never changes, so it could not detect a licence change in DevIL itself. The tarball ships its own LICENSE one level above S, so checksum that instead. The path is relative to S, the form oe-core uses for a licence outside S, and it carries no unexpanded variable. Validated on MACHINE=ls1046ardb: do_populate_lic re-ran and passed, so the md5 matches; the buildhistory packages tree is byte-identical. Signed-off-by: Luciano Dittgen

Merge pull request #2660 from lucianogdittgen/oelint-vars-outofcontext

2026-09-02T17:10:01+00:00

Set FILESEXTRAPATHS only where it is needed

imx-gpu-viv: Drop the dead FILESEXTRAPATHS entry

2026-09-02T11:43:18+00:00

The include adds ${THISDIR}/imx-gpu-viv to the file search path, but that directory no longer exists. Commit fc53cc24 ("imx-gpu-viv: Simplify using imx-gpu-viv-6.inc from another layer") added the entry in 2023 so a layer requiring this include could still find the SRC_URI file imx_icd.json; commit 610fda22 ("imx-gpu-viv: Fix the imx-gpu-viv install task") later dropped that file, and the directory went with it. Nothing resolves a file through the search path any more: SRC_URI holds a single .bin URL with no file:// entry, and the LIC_FILES_CHKSUM file://COPYING reference resolves inside ${S}, which the EULA unpack produces. The entry is also :append, so it sat at the end of FILESPATH and could not take priority over anything. Removing it leaves the search path unchanged. On imx8mp-lpddr4-evk, bitbake -e reports the same 60 unique directories before and after: the entry only duplicated paths that the default ${FILE_DIRNAME}/${BPN} already supplies, so FILESPATH shrinks from 84 entries to 63 while it reaches exactly the same set. None of those directories exists on disk. Validated with buildhistory: rebuilt imx-gpu-viv for MACHINE=imx8mp-lpddr4-evk with do_unpack forced, before and after. The buildhistory diff is empty apart from the metadata-revs "-- modified" marker, so no package, file list, size or dependency moved. Found by the oelint rule oelint.vars.outofcontext, which asks that FILESEXTRAPATHS be set only in a bbappend. The same edit also clears oelint.vars.fileextrapaths. Signed-off-by: Luciano Dittgen

Merge pull request #2650 from lucianogdittgen/oelint-vars-specific

2026-09-01T22:13:21+00:00

oelint: Resolve vars.specific on dynamic split packages via PACKAGES_DYNAMIC