| Message ID | a5b379d1-88b8-499c-bfe6-d741e508ca84@gmail.com |
|---|---|
| State | New |
| Headers | show |
| Series | rpm-sequoia-crypto-policy: add nativesdk to BBCLASSEXTEND | expand |
On Sun, 28 Sept 2025 at 22:32, Clayton Casciato via lists.openembedded.org <majortomtosourcecontrol=gmail.com@lists.openembedded.org> wrote: > > Prerequisite for nativesdk-rpm. > Would be required by nativesdk-rpm-sequoia > > ERROR: Nothing RPROVIDES 'nativesdk-rpm-sequoia-crypto-policy' > (but virtual:nativesdk: > [...]/meta/recipes-devtools/rpm-sequoia/rpm-sequoia_1.9.0.bb > RDEPENDS on or otherwise requires it) > > Reproduce: > 1. Add "sequoia" to rpm's default PACKAGECONFIG [1] > 2. Add "nativesdk" to rpm-sequoia's BBCLASSEXTEND > 3. Configure local.conf [2] > 4. Attempt to build "core-image-minimal" > > [1] This is seen when directly modifying the weak default. > This does not occur with "PACKAGECONFIG:append:pn-rpm". ... > -BBCLASSEXTEND = "native" > +BBCLASSEXTEND = "native nativesdk" There has to be an actual use case for doing this, and some standard build path where nativesdk variant is tested. Otherwise, it's a workaround, and you need to add sequoia option such that it's enabled only for native and target. Alex
>> On Sun, 28 Sept 2025 at 22:32, Clayton Casciato via >> lists.openembedded.org >> <majortomtosourcecontrol=gmail.com@lists.openembedded.org> wrote: >>> >>> This does not occur with "PACKAGECONFIG:append:pn-rpm". >>@ >> ... >>@ >>> -BBCLASSEXTEND = "native" >>> +BBCLASSEXTEND = "native nativesdk" >>@ >> There has to be an actual use case for doing this, and some standard >> build path where nativesdk variant is tested. Otherwise, it's a >> workaround, and you need to add sequoia option such that it's enabled >> only for native and target. >>@ >> Alex >@ Hi, Alex I understand the pushback against unnecessary functionality. I would still like to resolve the case where "sequoia" is in the ??. What are your thoughts on this alternative? PACKAGECONFIG:remove:class-nativesdk = "sequoia" Thank you! Clayton Casciato
On Mon, 29 Sept 2025 at 15:47, Clayton Casciato <majortomtosourcecontrol@gmail.com> wrote: > I understand the pushback against unnecessary functionality. > > I would still like to resolve the case where "sequoia" is in the ??. > > What are your thoughts on this alternative? > > PACKAGECONFIG:remove:class-nativesdk = "sequoia" I would rather add it separately like the signing selftest does: meta/lib/oeqa/selftest/cases/signing.py: feature += 'PACKAGECONFIG:append:pn-rpm-native = " sequoia"\n' meta/lib/oeqa/selftest/cases/signing.py: feature += 'PACKAGECONFIG:append:pn-rpm = " sequoia"\n' _remove works 'after the fact' and complicates understanding how the configuration ends up with the final values. It also cannot be undone with any other statement. Alex
> On Mon, 29 Sept 2025 at 15:47, Clayton Casciato > <majortomtosourcecontrol@gmail.com> wrote: >> I understand the pushback against unnecessary functionality. >> >> I would still like to resolve the case where "sequoia" is in the ??. >> >> What are your thoughts on this alternative? >> >> PACKAGECONFIG:remove:class-nativesdk = "sequoia" >@ > I would rather add it separately like the signing selftest does: >@ > meta/lib/oeqa/selftest/cases/signing.py: feature += > 'PACKAGECONFIG:append:pn-rpm-native = " sequoia"\n' > meta/lib/oeqa/selftest/cases/signing.py: feature += > 'PACKAGECONFIG:append:pn-rpm = " sequoia"\n' >@ > _remove works 'after the fact' and complicates understanding how the > configuration ends up with the final values. It also cannot be undone > with any other statement. >@ > Alex I will reconsider for the future, but I typically manipulate the non-specific PACKAGECONFIG as suggested by the documentation: https://docs.yoctoproject.org/dev/ref-manual/variables.html#term-PACKAGECONFIG This was a frustrating issue to encounter as I had not come across this behavior before for other recipes. I see the downside to using a remove operation. With that said, it is discoverable via bitbake-getvar. While this does seem like a bug, I plan to drop this issue for now. Thanks again! Clayton Casciato
diff --git a/meta/recipes-devtools/rpm-sequoia/rpm-sequoia-crypto-policy_git.bb b/meta/recipes-devtools/rpm-sequoia/rpm-sequoia-crypto-policy_git.bb index 19b86ae391..c602bb4b5a 100644 --- a/meta/recipes-devtools/rpm-sequoia/rpm-sequoia-crypto-policy_git.bb +++ b/meta/recipes-devtools/rpm-sequoia/rpm-sequoia-crypto-policy_git.bb @@ -34,4 +34,4 @@ do_install () { FILES:${PN} = "${datadir}/crypto-policies/back-ends/*" -BBCLASSEXTEND = "native" +BBCLASSEXTEND = "native nativesdk"
Prerequisite for nativesdk-rpm. Would be required by nativesdk-rpm-sequoia ERROR: Nothing RPROVIDES 'nativesdk-rpm-sequoia-crypto-policy' (but virtual:nativesdk: [...]/meta/recipes-devtools/rpm-sequoia/rpm-sequoia_1.9.0.bb RDEPENDS on or otherwise requires it) Reproduce: 1. Add "sequoia" to rpm's default PACKAGECONFIG [1] 2. Add "nativesdk" to rpm-sequoia's BBCLASSEXTEND 3. Configure local.conf [2] 4. Attempt to build "core-image-minimal" [1] This is seen when directly modifying the weak default. This does not occur with "PACKAGECONFIG:append:pn-rpm". [2] INHERIT += "sign_rpm" RPM_GPG_NAME = "[...]" RPM_GPG_PASSPHRASE = "[...]" GPG_PATH = "[...]" Signed-off-by: Clayton Casciato <majortomtosourcecontrol@gmail.com> --- .../rpm-sequoia/rpm-sequoia-crypto-policy_git.bb | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)