Re: [PATCH 0/1] image: introduce IMAGE_INITRD, deprecate INITRD_IMAGE

[Zhihang Wei <wzh@ilbers.de>]

Hi Cedric,

After some checking I think we need to rework this patch.

The INITRD_IMAGE variable was origionally used, when we already have a 
customized initrd image binary. In this case, the default initrd image 
should not be generated.
And the logic was implemented as (until commit ea6f44c9):
In image bb.class:
- If 'INITRD_IMAGE' is not empty, set 'no-generate-initrd'.
- Since image.bbclass inherits rootfs.bbclass:
   - rootfs.bbclass checks 'no-generate-initrd'.
   - If it is NOT set, add the task to generate the default initrd.

However, this did not actually work. the default initrd image was always 
generated.
The issue came from initramfs.bbclass:
- initramfs.bbclass has no logic to set 'no-generate-initrd'.
- It also inherits rootfs.bbclass, which means:
   - rootfs.bbclass sees 'no-generate-initrd' as unset, so it always 
generates default initrd image.

The patch "[v3 1/1] rootfs: rework no-generate-initrd rootfs feature" 
fixes this behavior, and it was merged to next as 78d42de0. However, I 
tested your patch before rebasing onto the latest next. It passed CI and 
your additional test cases, and I applied it as c3c4e72c. At that moment 
it caused a merge conflict, and I did not resolve the conflict correctly.

After both the patches were merged to next, two of the test cases you 
added failed:
- InitRdTest.test_var_initrd_image, and
- InitRdTest.test_var_image_initrd_and_initrd_image.
Both tests set INITRD_IMAGE variable to a meaningful value. But the 
logic for determining whether to generate a default initrd has been 
reworked and moved entirely into rootfs.bbclass. Under the new logic:
- rootfs.bbclass does not add 'generate-initrd' when INITRD_IMAGE is set.
- rootfs.bbclass adds initrd generation tasks only when 
'generate-initrd' is set.
- Since both initramfs.bbclass and image.bbclass inherit rootfs.bbclass, 
neither will generate a default initrd image in this case.

Back to your patch, I'm wondering whether it makes sense to keep both 
INITRD_IMAGE and your new variable. If my understanding is correct, they 
serve different purposes:
- INITRD_IMAGE is used when a custom initrd image binary is provided.
- Your variable is used when a custom recipe should be used to build the 
initrd image.

Because I did not resolve the merge conflict correctly when applying 
your patch, I suggest reverting c3c4e72c for now and reworking your 
patch on top of 78d42de0.

Thanks,
Zhihang

On 11/12/25 17:41, Zhihang Wei wrote:
> Applied to next, thanks.
>
> The patch was rebased onto the current next.
>
> Best regards,
> Zhihang
>
> On 11/2/25 20:01, 'Cedric Hombourger' via isar-users wrote:
>> INITRD_IMAGE could be set to the full name of an initrd image to be
>> found in DEPLOY_DIR_IMAGE. Downstream layers had to assume how Isar
>> names its image artifacts (${DISTRO}-${IMAGE}-initrd.img suffix for
>> initrd images) and insert the build of their initramfs recipe into
>> the image build pipeline.
>>
>> IMAGE_INITRD is introduced to (1) have a constistent naming convention
>> for variables consumed by the image class (IMAGE_ prefix) (2) only
>> require this variable to be set when a custom initrd should be built
>> and used by the imager.
>>
>> For instance, cip-core was doing the following:
>>
>>     INITRAMFS_RECIPE ?= "cip-core-initramfs"
>>     INITRD_IMAGE = "${INITRAMFS_RECIPE}-${DISTRO}-${MACHINE}.initrd.img"
>>     do_image_wic[depends] += "${INITRAMFS_RECIPE}:do_build"
>>
>> And it could now be changed to:
>>
>>     IMAGE_INITRD ?= "cip-core-initramfs"
>>
>> (their layer would likely keep their INITRAMFS_RECIPE variable as it
>> may be considered as an API towards its own downstream users).
>>
>> INITRD_IMAGE is "only" deprecated; meaning that it may still be used
>> (but build-time warnings will be raised). If both IMAGE_INITRD and
>> INITRD_IMAGE are set then the latter will be ignored (a warning noting
>> that both were set will be emitted).
>>
>> These changes were tested against (1) new CI test-cases and (2) CIP
>> Core 1.9.0 both with and without adaptions to its code.
>>
>> Patch dependencies:
>>   [1] 
>> https://lists.isar-build.org/isar-users/20251101111326.600581-1-cedric.hombourger@siemens.com/T/#t
>>   [2] 
>> https://lists.isar-build.org/isar-users/20251030094451.1303871-1-Quirin.Gylstorff@siemens.com/T/#t
>>
>> InitRd tests:
>>
>>    builder@ed1e0b7e8d4e:/work/testsuite$ trun --max-parallel-tasks=1 
>> -p sstate=1 -t initrd citest.py
>>    JOB ID     : bdf8f7130f426014b0c008d0095ad317a9f66bee
>>    JOB LOG    : 
>> /tmp/tmp72bozbdi/avocado/job-results/job-2025-11-02T18.30-bdf8f71/job.log
>>     (1/8) citest.py:InitRdTest.test_dracut_in_image: STARTED
>>     (1/8) citest.py:InitRdTest.test_dracut_in_image: PASS (104.93 s)
>>     (2/8) citest.py:InitRdTest.test_dracut_build_initrd: STARTED
>>     (2/8) citest.py:InitRdTest.test_dracut_build_initrd: PASS (34.36 s)
>>     (3/8) citest.py:InitRdTest.test_dracut_build_failure: STARTED
>>     (3/8) citest.py:InitRdTest.test_dracut_build_failure: FAIL: 
>> Bitbake suceeded but was expected to fail! (143.54 s)
>>     (4/8) citest.py:InitRdTest.test_var_initrd_image: STARTED
>>     (4/8) citest.py:InitRdTest.test_var_initrd_image: PASS (259.88 s)
>>     (5/8) citest.py:InitRdTest.test_var_image_initrd: STARTED
>>     (5/8) citest.py:InitRdTest.test_var_image_initrd: PASS (237.34 s)
>>     (6/8) 
>> citest.py:InitRdTest.test_var_image_initrd_and_initrd_image: STARTED
>>     (6/8) 
>> citest.py:InitRdTest.test_var_image_initrd_and_initrd_image: PASS 
>> (242.03 s)
>>     (7/8) citest.py:InitRdCrossTests.test_dracut_in_image: STARTED
>>     (7/8) citest.py:InitRdCrossTests.test_dracut_in_image: PASS 
>> (407.83 s)
>>     (8/8) citest.py:InitRdCrossTests.test_dracut_build_initrd: STARTED
>>     (8/8) citest.py:InitRdCrossTests.test_dracut_build_initrd: PASS 
>> (44.71 s)
>>    RESULTS    : PASS 7 | ERROR 0 | FAIL 1 | SKIP 0 | WARN 0 | 
>> INTERRUPT 0 | CANCEL 0
>>    JOB TIME   : 1482.17 s
>>
>> See [1] and/or [2] for an explanation (and proposed fix) for the single
>> test failure.
>>
>> Cedric Hombourger (1):
>>    image: introduce IMAGE_INITRD, deprecate INITRD_IMAGE
>>
>>   RECIPE-API-CHANGELOG.md    | 28 +++++++++++++
>>   doc/user_manual.md         |  3 +-
>>   meta/classes/image.bbclass | 31 ++++++++++++--
>>   testsuite/citest.py        | 83 ++++++++++++++++++++++++++++++++++++++
>>   4 files changed, 140 insertions(+), 5 deletions(-)
>>
>

-- 
You received this message because you are subscribed to the Google Groups "isar-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to isar-users+unsubscribe@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/isar-users/5ae801ae-55e0-48ab-9db4-a55d7c03954b%40ilbers.de.