The PR build workflow does not need to be started every time a single
reviwer is added, since requesting a review does not change the build.
If the 'Build' label was already added earlier, the build workflow will
have been started already.
On every PR, a workflow is started to check if artifacts should be
built. This happens not only once, but many times, e.g. for every
selected reviewer. Since the workflow has `cancel-in-progress`
enabled, workflows are started and immediately cancelled by the
next one, resulting in many notifications.
Move the `cancel-in-progress` concurrency policy to the second job
which starts only after a check is done if the 'Build' label is even
active on the PR. This should greatly reduce "Workflow cancelled"
notifications via GitHub and email (if enabled by the user).
Also make wording more clear.
- add `pre_config_uboot_target` hook for switching BOOTCONFIG across the two targets
- adapt `UBOOT_TARGET_MAP` to not call the defconfig Makefile targets directly, instead, just do a variable assignment (ignored by Make)
- otherwise, when using the defconfig directly in the `UBOOT_TARGET_MAP`, the `post_config_uboot_target` .config changes are overwritten when Make is called
- only patch left is boot usb-nvme-scsi/sata first (still done in meson64.h)
- remove FIP handling from family file `meson-sm1.conf` into board file hook where it belongs
- u-boot: enable more compression methods, EFI debugging, i2c, leds, tcp networking
- use flashcp for mtd writing
- Unchanged:
- confirmed as of v2024.04: using the C4 (not HC4) defconfig is still needed to be able to write to mtd when booted from SD
- also confirmed: one still needs to erase Petitboot using Petitboot, then boot from SD, to be able to flash mainline u-boot to mtd
- exclusive for edge / Kwiboo:
- enable UMS / RockUSB and other gadget stuff
- enable TCP networking features, wget etc
- enable EFI debugging stuff
- store environment in SPI
- uses preboot to flash led briefly
- this code pattern is becoming copypasta fast and needs refactor before it spreads too much
- also for all branches (legacy/vendor/edge):
- nanopct6: updated blobs to DDR v1.16 and BL31 v1.45
- `git format-patch --zero-commit` doesn't affect `index xxx...yyy` lines, only `From: `
- so use the _classy_ "use a regex with a callback" solution as git format-patch doesn't offer one
- this will make _all_ patches change when rewritten, but hopefully _for the last time_ !
- we need to preserve `index 000000000000..xxx` as zeros, which indicate new file creation, thus:
- new file creations are rewritten as `index 000000000000..111111111111`
- non-creations are rewritten as `index 111111111111..222222222222`
- this is the final version of #6455
* rename .dts sources in overlay directory as *.dtso
* change "target +=" line in "dts-y +=" in overlay Makefile
* remove "always +=" line in overlay Makefile
* add .scr compilation support in kernel scripts/Makefile.lib
* patch kernel scripts/Makefile.dtbinst to avoid flattening overlay
directory
for the last two points, see
general-add-overlay-compilation-support.patch
- makes it compatible with vendor out-of-box blobs (which include TEE) in the from-factory eMMC
- Armbian itself doesn't ship TEE blobs
- when combining from-factory eMMC blobs with an Armbian SD card, TEE blobs are in practice found by u-boot
- but then proceeds to fail with `optee api revision is too low`
- disable OPTEE in defconfig fixes it, TEE isn't used in any way by Armbian
- makes it compatible with vendor out-of-box blobs (which include TEE) in the from-factory eMMC
- Armbian itself doesn't ship TEE blobs
- when combining from-factory eMMC blobs with an Armbian SD card, TEE blobs are in practice found by u-boot
- but then proceeds to fail with `optee api revision is too low`
- disable OPTEE in defconfig fixes it, TEE isn't used in any way by Armbian
