fix conflicting boot.scr if UBOOT_ENV is set #1310
Conversation
|
please also cherry-pick to kirkstone branch if good. thx. |
There was a problem hiding this comment.
Thanks for this! I think this makes sense. Could you reformat a bit the commit using https://meta-raspberrypi.readthedocs.io/en/latest/contributing.html?
When embedding meta-raspberry in a distro that uses UBOOT_ENV to ship their own custom boot.scr file, the current u-boot-default-script will make the build fail to build due to conflicts in $DEPLOY_DIR_IMAGE This patch fixes this by not adding u-boot-default-script to the u-boot DEPENDS if UBOOT_ENV is set. Signed-off-by: Alexander Sack <asac@pantacor.com>
asac
force-pushed
the
bugfix/prevent-conflict-boot-scr-from-UBOOT_ENV
branch
from
ba74562
to
11b1f99
Compare
|
OK @agherzan, I improved the commit message. From grepping the core code it seems most code that decides whether to install something or not is really hooked on UBOOT_ENV ... also UBOOT_ENV_SUFFIX has a default so it would not be an empty check. I suggest we land this as is and see if someone else runs into issues. |
| @@ -6,7 +6,7 @@ SRC_URI:append:rpi = " \ | |||
|
|
|||
| SRC_URI:append:rpi = " file://0001-rpi-always-set-fdt_addr-with-firmware-provided-FDT-address.patch" | |||
|
|
|||
| DEPENDS:append:rpi = " u-boot-default-script" | |||
| DEPENDS:append:rpi = " ${@oe.utils.conditional('UBOOT_ENV', '', 'u-boot-default-script', '', d)}" | |||
There was a problem hiding this comment.
I think its better to tweak u-boot-default-scripts recipe via a bbappend in such a distro. This is introducing a variable that wont be used elsewhere.
There was a problem hiding this comment.
Hey @kraj UBOOT_ENV is not a new variable invented by us. From what I read in the code it's the "official" core way of adding a boot.scr in a standardized manner ... https://github.com/openembedded/openembedded-core/blob/master/meta/classes-recipe/uboot-config.bbclass#L61 ... the u-boot_%.bbappend of meta-raspberrypi breaks that behaviour right now.
Maybe we should also check if template support for this in openembedded-core would be ok.
There was a problem hiding this comment.
right I did not check that the variable is already in core layer. However, this variable is relevant for deploy tasks. the logic is still twisted. IMO this change seems more appropriate in u-boot-defaults-script recipe for meta-rpi since this recipe provides the UBOOT_ENV expected file.
No description provided.