| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
|
|
|
|
| |
OE-Core master has switched compatibility requirements from nanbield
to scarthgap in preparation for the upcoming 5.0 release.
https://wiki.yoctoproject.org/wiki/Releases
Signed-off-by: Denys Dmytriyenko <denys@konsulko.com>
Acked-by: Chirag Shilwant <c-shilwant@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Since the introduction of TI_COMMON_DEPLOY, we've been seeing frequent
Pseudo errors [1] breaking image build during development, in particular
with partial rebuilds due to the externalsrc bbclass.
The relevant part of the error log seems to be the following:
ERROR: Task (/home/schifferm/Devel/src/ci-meta-tq-kirkstone/sources/meta
-tq/meta-dumpling/recipes-images/images/tq-image-generic-debug.bb:do
_rootfs) failed with exit code '1'
Pseudo log:
path mismatch [1 link]: ino 12215614 db '/home/schifferm/Devel/src/ci-me
ta-tq-kirkstone/build-mba64xx/deploy-ti/images/tqma64xxl-mbax4xxl/tq
-image-generic-debug-tqma64xxl-mbax4xxl-20231020091917.testdata.json
' req '/home/schifferm/Devel/src/ci-meta-tq-kirkstone/build-mba64xx/
tmp/work/tqma64xxl_mbax4xxl-tq-linux/tq-image-generic-debug/1.0-r0/r
ootfs/usr/share/stress-ng/example-jobs/device.job;65324e26'.
My understanding is that all deploy dirs should be ignored by Pseudo, as
openembedded-core/meta/bitbake.conf includes the whole of DEPLOY_DIR in
PSEUDO_IGNORE_PATHS. With the introduction of TI_COMMON_DEPLOY,
DEPLOY_DIR_IMAGE is not covered by that setting anymore, so we add it in
meta-ti to avoid the issue.
[1] https://wiki.yoctoproject.org/wiki/Pseudo_Abort
Signed-off-by: Matthias Schiffer <matthias.schiffer@ew.tq-group.com>
Acked-by: Denys Dmytriyenko <denys@konsulko.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
There are couple of k3r5 components that need to be built for the target
and they are self-contained and don't use glibc. Moreover, when building
a cross-compile toolchain for k3r5, there's no need to build glibc and
all its dependencies. While build-time speed up is nice, the main benefit
comes when packaging a nativesdk toolchain for distribution, reducing the
number of packages and the resulting size of the toolchain installer.
To avoid potential namespace conflict in package feeds between nativesdk
binary packages of Linux glibc and k3r5 baremetal toolchains, adjust the
package suffix accordingly.
This also requires a rather impactful change in the deployment structure,
which affects CIs and user experience. By default different TCLIBC builds
(e.g. glibc and baremetal in this case) are built in separate TMPDIRs to
avoid conflicts, especially on the nativesdk side. Also the default config
has DEPLOY_DIR residing under corresponding TMPDIR. Of course, this can be
changed by local.conf or <distro>.conf, which meta-ti-bsp has no control
over. Unlike multiconfig-image-packager example that can access deployed
artifacts across multiconfigs to package in an image, we use wic and its
IMAGE_BOOT_FILES list to package a final image. Unfortunately, wic cannot
access deployed artifacts across multiconfigs in separate TMPDIRs, as it
uses DEPLOY_DIR_IMAGE to collect the artifacts. To overcome this, we need
to set up a "common" deploy directory for images and sdks produced by
multiconfigs, which is located outside of TMPDIRs, set by TI_COMMON_DEPLOY
variable that can be changed or adjusted by distros as needed.
Signed-off-by: Denys Dmytriyenko <denys@konsulko.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
|
|
|
|
|
| |
OE-Core is switching layer compatibility to nanbield for the upcoming release.
Signed-off-by: Denys Dmytriyenko <denys@konsulko.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Meta-clang provides a bbappend for mesa to use clang to accelerate it's
rasterization with runtime code generation through the use of llvmpipe.
With the addition of mesa-pvr we no longer get this for free with the
blanket mesa bbappend on the devices that explicitly request mesa-pvr so
lets add a dynamic layer to append the same args meta-clang does.
Though the number of devices that would select pvr-mesa and still want
to use software rendering is small, it's not zero due to debug and
testing purposes.
Signed-off-by: Randolph Sapp <rs@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
There are some unobvious issues with adding PVR support to Mesa
via a bbappend:
1. We need to mark mesa package as machine-specific, due to
differences in builds between SGX, Rogue and software-rendering
2. We also need to then mark mesa package as providing safe
ABIs (EGL/GLES/GBM) in order for all generic dependent packages
to not be treated as machine-specific, allowing their re-use
across different machines of the same architecture
But doing the above alters the upstream mesa package and changes
its signatures even when not building for TI platforms, which is
a Yocto Project compliance violation.
In order to resolve this issue, convert Mesa bbappend, that adds
PVR support, into its own standalone alternative provider, called
mesa-pvr and allow selecting it with PREFERRED_PROVIDER settings.
Signed-off-by: Denys Dmytriyenko <denys@konsulko.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
For TI platforms with SGX and Rogue graphics we patch upstream Mesa with
PVR support and conditionally enable features based on virtual/gpudriver
preference. This makes Mesa package platform-specific and it contains the
corresponding PACKAGE_ARCH = "${MACHINE_ARCH}" line.
In order to get greater re-use from generic non-machine-speciofic packages
down the dependency chain, such as Wayland, Weston, Qt5 modules, etc. we
need to mark mesa recipe as still providing stable OpenGL ABIs, such as
EGL, GLES, GBM, etc. - add it to SIGGEN_EXCLUDERECIPES_ABISAFE list.
Signed-off-by: Denys Dmytriyenko <denys@konsulko.com>
Reviewed-by: Andrew Davis <afd@ti.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Neither of recipes nor their ABI is all that stable. OpenGL might be
slightly more stable, but that is not what these provide anymore.
The variable helps with re-use of generic packages down the dependency
tree, when they depend on a machine-specific package. SGX UM used to
provide different standard OpenGL ABIs (EGL, GLES, GBM) and hence was
listed here. Now those ABIs are provided by Mesa, so SGX is removed.
Signed-off-by: Andrew Davis <afd@ti.com>
Signed-off-by: Denys Dmytriyenko <denys@konsulko.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
|
|
|
|
|
|
| |
Rather than just support mickledore we should also support the other
layers that we still support to ensure that as many as different
versions we can.
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
|
|
|
|
|
|
|
| |
oe-core switched to mickedore [1]
[1] https://git.openembedded.org/openembedded-core/commit/?id=57239d66b933c4313cf331d35d13ec2d0661c38f
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
|
|
|
| |
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
|
|
|
| |
Signed-off-by: Denys Dmytriyenko <denys@konsulko.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|
|
Extra sub-layers will be created next and relevant content moved across them.
Signed-off-by: Denys Dmytriyenko <denys@konsulko.com>
Signed-off-by: Ryan Eatmon <reatmon@ti.com>
|