On Sat, Jan 3, 2026 at 11:44 PM ChenQi <[email protected]> wrote:
>
> Hi Bruce,
>
> Thanks a lot for your prompt response and efforts.
>
> The BB_GIT_SHALLOW_EXTRA_REFS is the only one that actually breaks build.
> Switching to the tag parameter looks like a good approach.
> In case you want to check if things work for that new approach:
> 1. Add in local.conf:
> BB_GENERATE_SHALLOW_TARBALLS = "1"
> BB_GIT_SHALLOW = "1"
> 2. bitbake cni -c fetch

I have updates to the hybrid fetcher (it was missing some
de-duplication logic that
is part of the go fetcher), and regenerations to all the recipes completed.

The recipes all have BB_GIT_SHALLOW = "1" now (for testing), and use tag=
when the discovery hasn't "unshallowed" a clone and the reference is a tag.

shallow=1 is still in the src_uri entries, as I mentioned it is an
indication of intent
that I'm going to leverage in the future. It simply means, the
go-mod-vcs process
thinks that this repository will work in shallow mode.

Cheers,

Bruce

>
> Regards,
> Qi
>
> On 12/30/25 20:33, Bruce Ashfield wrote:
>
>
>
> On Tue, Dec 30, 2025 at 12:11 AM Bruce Ashfield via lists.yoctoproject.org 
> <[email protected]> wrote:
>>
>>
>>
>> On Mon, Dec 29, 2025 at 11:23 PM Chen, Qi <[email protected]> wrote:
>>>
>>> Hi Bruce,
>>>
>>>
>>>
>>> The new go-mod-vcs recently introduced in meta-virt  has some some problems.
>>>
>>>
>>>
>>> The first and most significant one is the misuse of 
>>> BB_GIT_SHALLOW_EXTRA_REFS. It causes fetching failure when shallow tarball 
>>> generation is enabled.
>>
>>
>> I wouldn't call it a misuse, it is actually required to make the support work
>> and attempt to save some download space. There are thousands of dependencies
>> for some of the recipes and MANY of them have tags that aren't on branches, 
>> or
>> aren't at the tip of branches, so those refs are there to force a deep 
>> enough fetch
>> or to ensure that bitbake does the same "deshallowing"  as was done during
>> generation.
>>
>> I could probably switch it out for tag= in the different entries. I've made 
>> that
>> tweak and will run some tests.
>>
>>
>>>
>>> Use the following steps to reproduce the error:
>>>
>>> Add in local.conf:
>>>
>>> BB_GENERATE_SHALLOW_TARBALLS = "1"
>>>
>>> BB_GIT_SHALLOW = "1"
>>>
>>> bitbake cni -c fetch
>>>
>>>
>>>
>>> Error message: fatal: couldn't find remote ref v0.0.4
>>>
>>> Root cause: the repos specified in SRC_URI does not contain the required 
>>> extra refs specified by BB_GIT_SHALLOW_EXTRA_REFS.
>>
>> I run with that all the time, I've never seen it, those 
>> BB_GIT_SHALLOW_EXTRA_REFS are
>> not magic, they come directly from the repositories during discovery and are 
>> required to deshallow
>> a repository.
>>
>> My tests with the tag parameter may solve the issue. So stay tuned, I have 
>> that and several other
>> changes pending. But none of them will be until the end of next week once I 
>> am back in the office.
>>
>>>
>>>
>>>
>>> The second one is the duplicate entries in SRC_URI. Example: errdefs in 
>>> docker-compose.
>>>
>>>
>>
>>
>> It builds fine here, what issues are you seeing with it ? Are you using VCS
>> mode ? or hybrid ?
>
>
> There are multiple places that the vcs generator looks for (and removes 
> duplicates).
> I added another and am testing it. So once I fix the shallow clone logic, I 
> can push
> this as well.
>
> I was able to build both modes locally, so it is a cosmetic fix, but worth 
> doing.
>
> Cheers,
>
> Bruce
>
>
>>
>>
>> You can be assured that anything I pushed has been built many times, and has
>> been runtime tested. So the first information you need to provide is more 
>> details
>> about what you are building, so I can track down the differences.
>>
>>>
>>> The third one is the ‘shallow=1’ parameter in SRC_URI. The problem is that 
>>> git fetcher does not support such parameter.
>>>
>>> In fetch2/git.py: ud.shallow = d.getVar("BB_GIT_SHALLOW") == "1"
>>>
>>> So I guess these could be removed? And switching to use 
>>> BB_GIT_SHALLOW:pn-xxx = “1”?
>>
>>
>> No. I plan to add more support for that later, it just isn't ready yet. They 
>> will stay.
>>
>>>
>>>
>>> P.S.
>>>
>>> I sent a related patch to bitbake: 
>>> https://lists.openembedded.org/g/bitbake-devel/message/18656
>>>
>>> The problem in bitbake is revealed by the current SRC_URI settings in 
>>> meta-virt.
>>>
>>> If you get similar error, you can apply the above patch.
>>
>>
>> If I was going to see that error, I would have already ...
>>
>> Bruce
>>
>>
>>>
>>>
>>>
>>> Regards,
>>>
>>> Qi
>>
>>
>>
>> --
>> - Thou shalt not follow the NULL pointer, for chaos and madness await thee 
>> at its end
>> - "Use the force Harry" - Gandalf, Star Trek II
>>
>>
>> 
>>
>
>
> --
> - Thou shalt not follow the NULL pointer, for chaos and madness await thee at 
> its end
> - "Use the force Harry" - Gandalf, Star Trek II
>
>


-- 
- Thou shalt not follow the NULL pointer, for chaos and madness await
thee at its end
- "Use the force Harry" - Gandalf, Star Trek II
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#9514): 
https://lists.yoctoproject.org/g/meta-virtualization/message/9514
Mute This Topic: https://lists.yoctoproject.org/mt/116993898/21656
Group Owner: [email protected]
Unsubscribe: https://lists.yoctoproject.org/g/meta-virtualization/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to