[PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Khem Raj]

dnf does not work with opkg or dpkg/apt anyway

Signed-off-by: Khem Raj <raj.khem@gmail.com>
---
v2: Use PNBLACKLIST instead of anon python

 meta/recipes-devtools/dnf/dnf_4.2.2.bb        | 2 ++
 meta/recipes-devtools/libdnf/libdnf_0.28.1.bb | 1 +
 2 files changed, 3 insertions(+)

diff --git a/meta/recipes-devtools/dnf/dnf_4.2.2.bb b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
index f38167f1ad..220f1aabbd 100644
--- a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
+++ b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
@@ -84,3 +84,5 @@ SYSTEMD_SERVICE_${PN} = "dnf-makecache.service dnf-makecache.timer \
                          dnf-automatic-notifyonly.service dnf-automatic-notifyonly.timer \
 "
 SYSTEMD_AUTO_ENABLE ?= "disable"
+
+PNBLACKLIST[dnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES', 'package_rpm', '', 'does not build correctly without package_rpm in PACKAGE_CLASSES', d)}"
diff --git a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
index 882c435b32..49afa04812 100644
--- a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
+++ b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
@@ -26,4 +26,5 @@ EXTRA_OECMAKE_append_class-native = " -DWITH_GIR=OFF"
 EXTRA_OECMAKE_append_class-nativesdk = " -DWITH_GIR=OFF"
 
 BBCLASSEXTEND = "native nativesdk"
+PNBLACKLIST[libdnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES', 'package_rpm', '', 'does not build correctly without package_rpm in PACKAGE_CLASSES', d)}"
 
-- 
2.25.0

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Alexander Kanavin]

[-- Attachment #1: Type: text/plain, Size: 2175 bytes --]

I would still prefer to just split rpm packaging into binaries and
libraries, and avoid the need to blacklist and tweak recipes based on what
PACKAGE_CLASSES is set to altogether. That setting should not leak into
recipes, and should only matter from do_package onwards. Can you look into
the rpm recipe please?

Alex

On Wed, 12 Feb 2020 at 20:19, Khem Raj <raj.khem@gmail.com> wrote:

> dnf does not work with opkg or dpkg/apt anyway
>
> Signed-off-by: Khem Raj <raj.khem@gmail.com>
> ---
> v2: Use PNBLACKLIST instead of anon python
>
>  meta/recipes-devtools/dnf/dnf_4.2.2.bb        | 2 ++
>  meta/recipes-devtools/libdnf/libdnf_0.28.1.bb | 1 +
>  2 files changed, 3 insertions(+)
>
> diff --git a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> index f38167f1ad..220f1aabbd 100644
> --- a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> +++ b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> @@ -84,3 +84,5 @@ SYSTEMD_SERVICE_${PN} = "dnf-makecache.service
> dnf-makecache.timer \
>                           dnf-automatic-notifyonly.service
> dnf-automatic-notifyonly.timer \
>  "
>  SYSTEMD_AUTO_ENABLE ?= "disable"
> +
> +PNBLACKLIST[dnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES',
> 'package_rpm', '', 'does not build correctly without package_rpm in
> PACKAGE_CLASSES', d)}"
> diff --git a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> index 882c435b32..49afa04812 100644
> --- a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> +++ b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> @@ -26,4 +26,5 @@ EXTRA_OECMAKE_append_class-native = " -DWITH_GIR=OFF"
>  EXTRA_OECMAKE_append_class-nativesdk = " -DWITH_GIR=OFF"
>
>  BBCLASSEXTEND = "native nativesdk"
> +PNBLACKLIST[libdnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES',
> 'package_rpm', '', 'does not build correctly without package_rpm in
> PACKAGE_CLASSES', d)}"
>
> --
> 2.25.0
>
> --
> _______________________________________________
> Openembedded-core mailing list
> Openembedded-core@lists.openembedded.org
> http://lists.openembedded.org/mailman/listinfo/openembedded-core
>

[-- Attachment #2: Type: text/html, Size: 3643 bytes --]

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Khem Raj]

On Wed, Feb 12, 2020 at 2:00 PM Alexander Kanavin
<alex.kanavin@gmail.com> wrote:
>
> I would still prefer to just split rpm packaging into binaries and libraries, and avoid the need to blacklist and tweak recipes based on what PACKAGE_CLASSES is set to altogether. That setting should not leak into recipes, and should only matter from do_package onwards. Can you look into the rpm recipe please?
>

 dnf and dnf related packages expecting rpmdb support in libresolv. I
do not know dnf well enough to fix it and moreover it has no direct
use for my usecase either. I also don't see a point of entertaining
dnf for non-rpm backends. Its not intended for that either even if we
were to make dnf not want rpmdb what good will it do to build for
distros who do not use rpm. These changes do not change poky defaults
which use rpm as default. OE-core does not and I dont know of any
opkg user who also has rpmdb needed.

> Alex
>
> On Wed, 12 Feb 2020 at 20:19, Khem Raj <raj.khem@gmail.com> wrote:
>>
>> dnf does not work with opkg or dpkg/apt anyway
>>
>> Signed-off-by: Khem Raj <raj.khem@gmail.com>
>> ---
>> v2: Use PNBLACKLIST instead of anon python
>>
>>  meta/recipes-devtools/dnf/dnf_4.2.2.bb        | 2 ++
>>  meta/recipes-devtools/libdnf/libdnf_0.28.1.bb | 1 +
>>  2 files changed, 3 insertions(+)
>>
>> diff --git a/meta/recipes-devtools/dnf/dnf_4.2.2.bb b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> index f38167f1ad..220f1aabbd 100644
>> --- a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> +++ b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> @@ -84,3 +84,5 @@ SYSTEMD_SERVICE_${PN} = "dnf-makecache.service dnf-makecache.timer \
>>                           dnf-automatic-notifyonly.service dnf-automatic-notifyonly.timer \
>>  "
>>  SYSTEMD_AUTO_ENABLE ?= "disable"
>> +
>> +PNBLACKLIST[dnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES', 'package_rpm', '', 'does not build correctly without package_rpm in PACKAGE_CLASSES', d)}"
>> diff --git a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> index 882c435b32..49afa04812 100644
>> --- a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> +++ b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> @@ -26,4 +26,5 @@ EXTRA_OECMAKE_append_class-native = " -DWITH_GIR=OFF"
>>  EXTRA_OECMAKE_append_class-nativesdk = " -DWITH_GIR=OFF"
>>
>>  BBCLASSEXTEND = "native nativesdk"
>> +PNBLACKLIST[libdnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES', 'package_rpm', '', 'does not build correctly without package_rpm in PACKAGE_CLASSES', d)}"
>>
>> --
>> 2.25.0
>>
>> --
>> _______________________________________________
>> Openembedded-core mailing list
>> Openembedded-core@lists.openembedded.org
>> http://lists.openembedded.org/mailman/listinfo/openembedded-core

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Alexander Kanavin]

[-- Attachment #1: Type: text/plain, Size: 3206 bytes --]

But you do not need to fix or touch dnf at all. You only need to adjust rpm
packaging and split it in two parts instead of bundling everything into a
single package.

Alex

On Wed 12. Feb 2020 at 23.16, Khem Raj <raj.khem@gmail.com> wrote:

> On Wed, Feb 12, 2020 at 2:00 PM Alexander Kanavin
> <alex.kanavin@gmail.com> wrote:
> >
> > I would still prefer to just split rpm packaging into binaries and
> libraries, and avoid the need to blacklist and tweak recipes based on what
> PACKAGE_CLASSES is set to altogether. That setting should not leak into
> recipes, and should only matter from do_package onwards. Can you look into
> the rpm recipe please?
> >
>
>  dnf and dnf related packages expecting rpmdb support in libresolv. I
> do not know dnf well enough to fix it and moreover it has no direct
> use for my usecase either. I also don't see a point of entertaining
> dnf for non-rpm backends. Its not intended for that either even if we
> were to make dnf not want rpmdb what good will it do to build for
> distros who do not use rpm. These changes do not change poky defaults
> which use rpm as default. OE-core does not and I dont know of any
> opkg user who also has rpmdb needed.
>
> > Alex
> >
> > On Wed, 12 Feb 2020 at 20:19, Khem Raj <raj.khem@gmail.com> wrote:
> >>
> >> dnf does not work with opkg or dpkg/apt anyway
> >>
> >> Signed-off-by: Khem Raj <raj.khem@gmail.com>
> >> ---
> >> v2: Use PNBLACKLIST instead of anon python
> >>
> >>  meta/recipes-devtools/dnf/dnf_4.2.2.bb        | 2 ++
> >>  meta/recipes-devtools/libdnf/libdnf_0.28.1.bb | 1 +
> >>  2 files changed, 3 insertions(+)
> >>
> >> diff --git a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> >> index f38167f1ad..220f1aabbd 100644
> >> --- a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> >> +++ b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> >> @@ -84,3 +84,5 @@ SYSTEMD_SERVICE_${PN} = "dnf-makecache.service
> dnf-makecache.timer \
> >>                           dnf-automatic-notifyonly.service
> dnf-automatic-notifyonly.timer \
> >>  "
> >>  SYSTEMD_AUTO_ENABLE ?= "disable"
> >> +
> >> +PNBLACKLIST[dnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES',
> 'package_rpm', '', 'does not build correctly without package_rpm in
> PACKAGE_CLASSES', d)}"
> >> diff --git a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> >> index 882c435b32..49afa04812 100644
> >> --- a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> >> +++ b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> >> @@ -26,4 +26,5 @@ EXTRA_OECMAKE_append_class-native = " -DWITH_GIR=OFF"
> >>  EXTRA_OECMAKE_append_class-nativesdk = " -DWITH_GIR=OFF"
> >>
> >>  BBCLASSEXTEND = "native nativesdk"
> >> +PNBLACKLIST[libdnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES',
> 'package_rpm', '', 'does not build correctly without package_rpm in
> PACKAGE_CLASSES', d)}"
> >>
> >> --
> >> 2.25.0
> >>
> >> --
> >> _______________________________________________
> >> Openembedded-core mailing list
> >> Openembedded-core@lists.openembedded.org
> >> http://lists.openembedded.org/mailman/listinfo/openembedded-core
>

[-- Attachment #2: Type: text/html, Size: 5089 bytes --]

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Khem Raj]

On Wed, Feb 12, 2020 at 8:49 PM Alexander Kanavin
<alex.kanavin@gmail.com> wrote:
>
> But you do not need to fix or touch dnf at all. You only need to adjust rpm packaging and split it in two parts instead of bundling everything into a single package.
>

when we do world builds without package_rpm, dnf and related packages
get compiled which is unnecessary and moreover wrong and when we mark
libsolv to enable rpmdb only for package_rpm, this becomes even more evident.

Perhaps, I am missing what you intend to convey, in that case maybe you can send
code changes to support what you are intending here.

> Alex
>
> On Wed 12. Feb 2020 at 23.16, Khem Raj <raj.khem@gmail.com> wrote:
>>
>> On Wed, Feb 12, 2020 at 2:00 PM Alexander Kanavin
>> <alex.kanavin@gmail.com> wrote:
>> >
>> > I would still prefer to just split rpm packaging into binaries and libraries, and avoid the need to blacklist and tweak recipes based on what PACKAGE_CLASSES is set to altogether. That setting should not leak into recipes, and should only matter from do_package onwards. Can you look into the rpm recipe please?
>> >
>>
>>  dnf and dnf related packages expecting rpmdb support in libresolv. I
>> do not know dnf well enough to fix it and moreover it has no direct
>> use for my usecase either. I also don't see a point of entertaining
>> dnf for non-rpm backends. Its not intended for that either even if we
>> were to make dnf not want rpmdb what good will it do to build for
>> distros who do not use rpm. These changes do not change poky defaults
>> which use rpm as default. OE-core does not and I dont know of any
>> opkg user who also has rpmdb needed.
>>
>> > Alex
>> >
>> > On Wed, 12 Feb 2020 at 20:19, Khem Raj <raj.khem@gmail.com> wrote:
>> >>
>> >> dnf does not work with opkg or dpkg/apt anyway
>> >>
>> >> Signed-off-by: Khem Raj <raj.khem@gmail.com>
>> >> ---
>> >> v2: Use PNBLACKLIST instead of anon python
>> >>
>> >>  meta/recipes-devtools/dnf/dnf_4.2.2.bb        | 2 ++
>> >>  meta/recipes-devtools/libdnf/libdnf_0.28.1.bb | 1 +
>> >>  2 files changed, 3 insertions(+)
>> >>
>> >> diff --git a/meta/recipes-devtools/dnf/dnf_4.2.2.bb b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> >> index f38167f1ad..220f1aabbd 100644
>> >> --- a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> >> +++ b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> >> @@ -84,3 +84,5 @@ SYSTEMD_SERVICE_${PN} = "dnf-makecache.service dnf-makecache.timer \
>> >>                           dnf-automatic-notifyonly.service dnf-automatic-notifyonly.timer \
>> >>  "
>> >>  SYSTEMD_AUTO_ENABLE ?= "disable"
>> >> +
>> >> +PNBLACKLIST[dnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES', 'package_rpm', '', 'does not build correctly without package_rpm in PACKAGE_CLASSES', d)}"
>> >> diff --git a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> >> index 882c435b32..49afa04812 100644
>> >> --- a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> >> +++ b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> >> @@ -26,4 +26,5 @@ EXTRA_OECMAKE_append_class-native = " -DWITH_GIR=OFF"
>> >>  EXTRA_OECMAKE_append_class-nativesdk = " -DWITH_GIR=OFF"
>> >>
>> >>  BBCLASSEXTEND = "native nativesdk"
>> >> +PNBLACKLIST[libdnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES', 'package_rpm', '', 'does not build correctly without package_rpm in PACKAGE_CLASSES', d)}"
>> >>
>> >> --
>> >> 2.25.0
>> >>
>> >> --
>> >> _______________________________________________
>> >> Openembedded-core mailing list
>> >> Openembedded-core@lists.openembedded.org
>> >> http://lists.openembedded.org/mailman/listinfo/openembedded-core

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Alexander Kanavin]

[-- Attachment #1: Type: text/plain, Size: 4595 bytes --]

World builds also enable apt and dpkg, but there is no problem with that
for you? There is no consistency here; if you want to start excluding
things from world based on what packaging is enabled, you need to treat all
options equally, including deb and ipk.

The original problem was that libsolv pulls in the full set of rpm
executables and libraries, which erroneously triggers the rpm test on
target. We can solve this by making libsolv pull in only the rpm libraries
and not executables, so that rpm test will be skipped. Or maybe even make
the test itself check whether rpm is the packaging format and skip
otherwise.

Alex

On Thu 13. Feb 2020 at 7.01, Khem Raj <raj.khem@gmail.com> wrote:

> On Wed, Feb 12, 2020 at 8:49 PM Alexander Kanavin
> <alex.kanavin@gmail.com> wrote:
> >
> > But you do not need to fix or touch dnf at all. You only need to adjust
> rpm packaging and split it in two parts instead of bundling everything into
> a single package.
> >
>
> when we do world builds without package_rpm, dnf and related packages
> get compiled which is unnecessary and moreover wrong and when we mark
> libsolv to enable rpmdb only for package_rpm, this becomes even more
> evident.
>
> Perhaps, I am missing what you intend to convey, in that case maybe you
> can send
> code changes to support what you are intending here.
>
> > Alex
> >
> > On Wed 12. Feb 2020 at 23.16, Khem Raj <raj.khem@gmail.com> wrote:
> >>
> >> On Wed, Feb 12, 2020 at 2:00 PM Alexander Kanavin
> >> <alex.kanavin@gmail.com> wrote:
> >> >
> >> > I would still prefer to just split rpm packaging into binaries and
> libraries, and avoid the need to blacklist and tweak recipes based on what
> PACKAGE_CLASSES is set to altogether. That setting should not leak into
> recipes, and should only matter from do_package onwards. Can you look into
> the rpm recipe please?
> >> >
> >>
> >>  dnf and dnf related packages expecting rpmdb support in libresolv. I
> >> do not know dnf well enough to fix it and moreover it has no direct
> >> use for my usecase either. I also don't see a point of entertaining
> >> dnf for non-rpm backends. Its not intended for that either even if we
> >> were to make dnf not want rpmdb what good will it do to build for
> >> distros who do not use rpm. These changes do not change poky defaults
> >> which use rpm as default. OE-core does not and I dont know of any
> >> opkg user who also has rpmdb needed.
> >>
> >> > Alex
> >> >
> >> > On Wed, 12 Feb 2020 at 20:19, Khem Raj <raj.khem@gmail.com> wrote:
> >> >>
> >> >> dnf does not work with opkg or dpkg/apt anyway
> >> >>
> >> >> Signed-off-by: Khem Raj <raj.khem@gmail.com>
> >> >> ---
> >> >> v2: Use PNBLACKLIST instead of anon python
> >> >>
> >> >>  meta/recipes-devtools/dnf/dnf_4.2.2.bb        | 2 ++
> >> >>  meta/recipes-devtools/libdnf/libdnf_0.28.1.bb | 1 +
> >> >>  2 files changed, 3 insertions(+)
> >> >>
> >> >> diff --git a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> >> >> index f38167f1ad..220f1aabbd 100644
> >> >> --- a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> >> >> +++ b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
> >> >> @@ -84,3 +84,5 @@ SYSTEMD_SERVICE_${PN} = "dnf-makecache.service
> dnf-makecache.timer \
> >> >>                           dnf-automatic-notifyonly.service
> dnf-automatic-notifyonly.timer \
> >> >>  "
> >> >>  SYSTEMD_AUTO_ENABLE ?= "disable"
> >> >> +
> >> >> +PNBLACKLIST[dnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES',
> 'package_rpm', '', 'does not build correctly without package_rpm in
> PACKAGE_CLASSES', d)}"
> >> >> diff --git a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> >> >> index 882c435b32..49afa04812 100644
> >> >> --- a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> >> >> +++ b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
> >> >> @@ -26,4 +26,5 @@ EXTRA_OECMAKE_append_class-native = "
> -DWITH_GIR=OFF"
> >> >>  EXTRA_OECMAKE_append_class-nativesdk = " -DWITH_GIR=OFF"
> >> >>
> >> >>  BBCLASSEXTEND = "native nativesdk"
> >> >> +PNBLACKLIST[libdnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES',
> 'package_rpm', '', 'does not build correctly without package_rpm in
> PACKAGE_CLASSES', d)}"
> >> >>
> >> >> --
> >> >> 2.25.0
> >> >>
> >> >> --
> >> >> _______________________________________________
> >> >> Openembedded-core mailing list
> >> >> Openembedded-core@lists.openembedded.org
> >> >> http://lists.openembedded.org/mailman/listinfo/openembedded-core
>

[-- Attachment #2: Type: text/html, Size: 7030 bytes --]

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Alexander Kanavin]

[-- Attachment #1: Type: text/plain, Size: 5719 bytes --]

I took a look at meta/lib/oeqa/runtime/cases/rpm.py.

All tests depend on test_rpm_query, which should be skipped if rpm is not
used for packaging:

        status, output = self.target.run('ls /var/lib/rpm/')
        if status != 0:
            self.skipTest('No /var/lib/rpm on target')

Can you please check, how the tests fails for you exactly, and why
/var/lib/rpm exists (and what is the content of it)?

The test used to have the check I am talking about, but it was removed by

commit 633bd85270922d71ab4b1f94cd166da51fbe1458
Author: Richard Purdie <richard.purdie@linuxfoundation.org>
Date:   Thu Nov 8 19:55:17 2018 +0000

    oeqa/runtime/cases: Improve test dependency information
...
-    @classmethod
-    def setUpClass(cls):
-        if cls.tc.td['PACKAGE_CLASSES'].split()[0] != 'package_rpm':
-            cls.skipTest('Tests require image to be build from rpm')

Alex


On Thu, 13 Feb 2020 at 08:53, Alexander Kanavin <alex.kanavin@gmail.com>
wrote:

> World builds also enable apt and dpkg, but there is no problem with that
> for you? There is no consistency here; if you want to start excluding
> things from world based on what packaging is enabled, you need to treat all
> options equally, including deb and ipk.
>
> The original problem was that libsolv pulls in the full set of rpm
> executables and libraries, which erroneously triggers the rpm test on
> target. We can solve this by making libsolv pull in only the rpm libraries
> and not executables, so that rpm test will be skipped. Or maybe even make
> the test itself check whether rpm is the packaging format and skip
> otherwise.
>
> Alex
>
> On Thu 13. Feb 2020 at 7.01, Khem Raj <raj.khem@gmail.com> wrote:
>
>> On Wed, Feb 12, 2020 at 8:49 PM Alexander Kanavin
>> <alex.kanavin@gmail.com> wrote:
>> >
>> > But you do not need to fix or touch dnf at all. You only need to adjust
>> rpm packaging and split it in two parts instead of bundling everything into
>> a single package.
>> >
>>
>> when we do world builds without package_rpm, dnf and related packages
>> get compiled which is unnecessary and moreover wrong and when we mark
>> libsolv to enable rpmdb only for package_rpm, this becomes even more
>> evident.
>>
>> Perhaps, I am missing what you intend to convey, in that case maybe you
>> can send
>> code changes to support what you are intending here.
>>
>> > Alex
>> >
>> > On Wed 12. Feb 2020 at 23.16, Khem Raj <raj.khem@gmail.com> wrote:
>> >>
>> >> On Wed, Feb 12, 2020 at 2:00 PM Alexander Kanavin
>> >> <alex.kanavin@gmail.com> wrote:
>> >> >
>> >> > I would still prefer to just split rpm packaging into binaries and
>> libraries, and avoid the need to blacklist and tweak recipes based on what
>> PACKAGE_CLASSES is set to altogether. That setting should not leak into
>> recipes, and should only matter from do_package onwards. Can you look into
>> the rpm recipe please?
>> >> >
>> >>
>> >>  dnf and dnf related packages expecting rpmdb support in libresolv. I
>> >> do not know dnf well enough to fix it and moreover it has no direct
>> >> use for my usecase either. I also don't see a point of entertaining
>> >> dnf for non-rpm backends. Its not intended for that either even if we
>> >> were to make dnf not want rpmdb what good will it do to build for
>> >> distros who do not use rpm. These changes do not change poky defaults
>> >> which use rpm as default. OE-core does not and I dont know of any
>> >> opkg user who also has rpmdb needed.
>> >>
>> >> > Alex
>> >> >
>> >> > On Wed, 12 Feb 2020 at 20:19, Khem Raj <raj.khem@gmail.com> wrote:
>> >> >>
>> >> >> dnf does not work with opkg or dpkg/apt anyway
>> >> >>
>> >> >> Signed-off-by: Khem Raj <raj.khem@gmail.com>
>> >> >> ---
>> >> >> v2: Use PNBLACKLIST instead of anon python
>> >> >>
>> >> >>  meta/recipes-devtools/dnf/dnf_4.2.2.bb        | 2 ++
>> >> >>  meta/recipes-devtools/libdnf/libdnf_0.28.1.bb | 1 +
>> >> >>  2 files changed, 3 insertions(+)
>> >> >>
>> >> >> diff --git a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> >> >> index f38167f1ad..220f1aabbd 100644
>> >> >> --- a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> >> >> +++ b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> >> >> @@ -84,3 +84,5 @@ SYSTEMD_SERVICE_${PN} = "dnf-makecache.service
>> dnf-makecache.timer \
>> >> >>                           dnf-automatic-notifyonly.service
>> dnf-automatic-notifyonly.timer \
>> >> >>  "
>> >> >>  SYSTEMD_AUTO_ENABLE ?= "disable"
>> >> >> +
>> >> >> +PNBLACKLIST[dnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES',
>> 'package_rpm', '', 'does not build correctly without package_rpm in
>> PACKAGE_CLASSES', d)}"
>> >> >> diff --git a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> >> >> index 882c435b32..49afa04812 100644
>> >> >> --- a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> >> >> +++ b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> >> >> @@ -26,4 +26,5 @@ EXTRA_OECMAKE_append_class-native = "
>> -DWITH_GIR=OFF"
>> >> >>  EXTRA_OECMAKE_append_class-nativesdk = " -DWITH_GIR=OFF"
>> >> >>
>> >> >>  BBCLASSEXTEND = "native nativesdk"
>> >> >> +PNBLACKLIST[libdnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES',
>> 'package_rpm', '', 'does not build correctly without package_rpm in
>> PACKAGE_CLASSES', d)}"
>> >> >>
>> >> >> --
>> >> >> 2.25.0
>> >> >>
>> >> >> --
>> >> >> _______________________________________________
>> >> >> Openembedded-core mailing list
>> >> >> Openembedded-core@lists.openembedded.org
>> >> >> http://lists.openembedded.org/mailman/listinfo/openembedded-core
>>
>

[-- Attachment #2: Type: text/html, Size: 8714 bytes --]

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 5558 bytes --]

On Wed, Feb 12, 2020 at 11:53 PM Alexander Kanavin <alex.kanavin@gmail.com>
wrote:

> World builds also enable apt and dpkg, but there is no problem with that
> for you? There is no consistency here; if you want to start excluding
> things from world based on what packaging is enabled, you need to treat all
> options equally, including deb and ipk.
>

No, problem is that if One wants to use opkg backend then rpm drags along
but apt or dpkg don’t
World builds are also problem but secondary
Please try to understand the entirety of problem

I tried everything to keep it building but i could not and I have been
saying that all along

If you can come up with patches to dnf to build with libresolv that is not
build with rpm support we might get to built however I don’t think it will
of any use at runtime


>
> The original problem was that libsolv pulls in the full set of rpm
> executables and libraries, which erroneously triggers the rpm test on
> target. We can solve this by making libsolv pull in only the rpm libraries
> and not executables,
>

Libsolve already has option to disable rpmdb support and we use it

so that rpm test will be skipped. Or maybe even make the test itself check
> whether rpm is the packaging format and skip otherwise.
>

No. Skipping test is just symptom of the problem I knew we could paper over
it but that’s not the point

Even if test is made to pass rpm will be dragged into
Images which is not desired


> Alex
>
> On Thu 13. Feb 2020 at 7.01, Khem Raj <raj.khem@gmail.com> wrote:
>
>> On Wed, Feb 12, 2020 at 8:49 PM Alexander Kanavin
>> <alex.kanavin@gmail.com> wrote:
>> >
>> > But you do not need to fix or touch dnf at all. You only need to adjust
>> rpm packaging and split it in two parts instead of bundling everything into
>> a single package.
>> >
>>
>> when we do world builds without package_rpm, dnf and related packages
>> get compiled which is unnecessary and moreover wrong and when we mark
>> libsolv to enable rpmdb only for package_rpm, this becomes even more
>> evident.
>>
>> Perhaps, I am missing what you intend to convey, in that case maybe you
>> can send
>> code changes to support what you are intending here.
>>
>> > Alex
>> >
>> > On Wed 12. Feb 2020 at 23.16, Khem Raj <raj.khem@gmail.com> wrote:
>> >>
>> >> On Wed, Feb 12, 2020 at 2:00 PM Alexander Kanavin
>> >> <alex.kanavin@gmail.com> wrote:
>> >> >
>> >> > I would still prefer to just split rpm packaging into binaries and
>> libraries, and avoid the need to blacklist and tweak recipes based on what
>> PACKAGE_CLASSES is set to altogether. That setting should not leak into
>> recipes, and should only matter from do_package onwards. Can you look into
>> the rpm recipe please?
>> >> >
>> >>
>> >>  dnf and dnf related packages expecting rpmdb support in libresolv. I
>> >> do not know dnf well enough to fix it and moreover it has no direct
>> >> use for my usecase either. I also don't see a point of entertaining
>> >> dnf for non-rpm backends. Its not intended for that either even if we
>> >> were to make dnf not want rpmdb what good will it do to build for
>> >> distros who do not use rpm. These changes do not change poky defaults
>> >> which use rpm as default. OE-core does not and I dont know of any
>> >> opkg user who also has rpmdb needed.
>> >>
>> >> > Alex
>> >> >
>> >> > On Wed, 12 Feb 2020 at 20:19, Khem Raj <raj.khem@gmail.com> wrote:
>> >> >>
>> >> >> dnf does not work with opkg or dpkg/apt anyway
>> >> >>
>> >> >> Signed-off-by: Khem Raj <raj.khem@gmail.com>
>> >> >> ---
>> >> >> v2: Use PNBLACKLIST instead of anon python
>> >> >>
>> >> >>  meta/recipes-devtools/dnf/dnf_4.2.2.bb        | 2 ++
>> >> >>  meta/recipes-devtools/libdnf/libdnf_0.28.1.bb | 1 +
>> >> >>  2 files changed, 3 insertions(+)
>> >> >>
>> >> >> diff --git a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> >> >> index f38167f1ad..220f1aabbd 100644
>> >> >> --- a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> >> >> +++ b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
>> >> >> @@ -84,3 +84,5 @@ SYSTEMD_SERVICE_${PN} = "dnf-makecache.service
>> dnf-makecache.timer \
>> >> >>                           dnf-automatic-notifyonly.service
>> dnf-automatic-notifyonly.timer \
>> >> >>  "
>> >> >>  SYSTEMD_AUTO_ENABLE ?= "disable"
>> >> >> +
>> >> >> +PNBLACKLIST[dnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES',
>> 'package_rpm', '', 'does not build correctly without package_rpm in
>> PACKAGE_CLASSES', d)}"
>> >> >> diff --git a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> >> >> index 882c435b32..49afa04812 100644
>> >> >> --- a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> >> >> +++ b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
>> >> >> @@ -26,4 +26,5 @@ EXTRA_OECMAKE_append_class-native = "
>> -DWITH_GIR=OFF"
>> >> >>  EXTRA_OECMAKE_append_class-nativesdk = " -DWITH_GIR=OFF"
>> >> >>
>> >> >>  BBCLASSEXTEND = "native nativesdk"
>> >> >> +PNBLACKLIST[libdnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES',
>> 'package_rpm', '', 'does not build correctly without package_rpm in
>> PACKAGE_CLASSES', d)}"
>> >> >>
>> >> >> --
>> >> >> 2.25.0
>> >> >>
>> >> >> --
>> >> >> _______________________________________________
>> >> >> Openembedded-core mailing list
>> >> >> Openembedded-core@lists.openembedded.org
>> >> >> http://lists.openembedded.org/mailman/listinfo/openembedded-core
>>
>

[-- Attachment #2: Type: text/html, Size: 8960 bytes --]

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Alexander Kanavin]

[-- Attachment #1: Type: text/plain, Size: 462 bytes --]

On Thu, 13 Feb 2020 at 17:06, Khem Raj <raj.khem@gmail.com> wrote:

> No, problem is that if One wants to use opkg backend then rpm drags along
> but apt or dpkg don’t
>
> Even if test is made to pass rpm will be dragged into
> Images which is not desired
>

We can modify the rpm packaging so that only the library from rpm that
libresolv links with, but not the executables or other libraries are in the
image. Would you be ok with that?

Alex

[-- Attachment #2: Type: text/html, Size: 910 bytes --]

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 942 bytes --]

On Thu, Feb 13, 2020 at 8:24 AM Alexander Kanavin <alex.kanavin@gmail.com>
wrote:

> On Thu, 13 Feb 2020 at 17:06, Khem Raj <raj.khem@gmail.com> wrote:
>
>> No, problem is that if One wants to use opkg backend then rpm drags along
>> but apt or dpkg don’t
>>
>> Even if test is made to pass rpm will be dragged into
>> Images which is not desired
>>
>
> We can modify the rpm packaging so that only the library from rpm that
> libresolv links with, but not the executables or other libraries are in the
> image. Would you be ok with that?
>

This still sneaks in pieces of rpm into distros where rpm is not needed so
better would be to retool dnf to not demand rpm support in libsolv and
break the dependency chain. Opkg is usually sought for small systems and
last thing we want is unneeded dependencies

I would rather use the OE configuration powers as such but I am not opposed
to such a solution


> Alex
>

[-- Attachment #2: Type: text/html, Size: 1937 bytes --]

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Alexander Kanavin]

[-- Attachment #1: Type: text/plain, Size: 692 bytes --]

On Thu, 13 Feb 2020 at 17:42, Khem Raj <raj.khem@gmail.com> wrote:

> This still sneaks in pieces of rpm into distros where rpm is not needed so
> better would be to retool dnf to not demand rpm support in libsolv and
> break the dependency chain. Opkg is usually sought for small systems and
> last thing we want is unneeded dependencies
>
> I would rather use the OE configuration powers as such but I am not
> opposed to such a solution
>

You can always set these things (rpm disabling in libsolv, blacklisting
libdnf/dnf) in your distro configuration, but I would like less complexity
and special-casing in poky, not more, for the pieces that I maintain at
least.

Alex

[-- Attachment #2: Type: text/html, Size: 1113 bytes --]

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Khem Raj]

[-- Attachment #1: Type: text/plain, Size: 1238 bytes --]

On Thu, Feb 13, 2020 at 8:57 AM Alexander Kanavin <alex.kanavin@gmail.com>
wrote:

> On Thu, 13 Feb 2020 at 17:42, Khem Raj <raj.khem@gmail.com> wrote:
>
>> This still sneaks in pieces of rpm into distros where rpm is not needed
>> so better would be to retool dnf to not demand rpm support in libsolv and
>> break the dependency chain. Opkg is usually sought for small systems and
>> last thing we want is unneeded dependencies
>>
>> I would rather use the OE configuration powers as such but I am not
>> opposed to such a solution
>>
>
> You can always set these things (rpm disabling in libsolv, blacklisting
> libdnf/dnf) in your distro configuration, but I would like less complexity
> and special-casing in poky, not more, for the pieces that I maintain at
> least.
>

Btw it’s being proposed for Oe-core not poky and in Oe core rpm is not
default backend but opkg is
Moreover it is being done in a way where poky default are being taken into
account

I really fail to understand your line of arguments
Here I am sorry, there seems to be no technical arguments for alternatives
but for convenience

I certainly can this goes unsaid and it just takes distro one step away
from Oe-core


> Alex
>

[-- Attachment #2: Type: text/html, Size: 2288 bytes --]

Re: [PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Alexander Kanavin]

[-- Attachment #1: Type: text/plain, Size: 452 bytes --]

On Thu, 20 Feb 2020 at 18:25, Khem Raj <raj.khem@gmail.com> wrote:

> +PNBLACKLIST[dnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES',
> 'package_rpm', '', 'does not build correctly without package_rpm in
> PACKAGE_CLASSES', d)}"
>

I meant that this string is extended:
"does not build correctly without package_rpm in PACKAGE_CLASSES due to
disabled rpm support in libsolv"

Otherwise it's vague and doesn't provide the real reason.

Alex

[-- Attachment #2: Type: text/html, Size: 856 bytes --]

[PATCH V2] dnf, libdnf: Ignore if PACKAGE_CLASSES does not have rpm

[Khem Raj]

dnf depends on libsolv, and libsolv does not have and easy way to
jusr enable rpmdb support, it either has full rpm or nothing, perhaps it
could be improved in libsolv and we can use that but from now we have
no other option

dnf support only rpm backend and wont work with opkg or dpkg/apt anyway

Signed-off-by: Khem Raj <raj.khem@gmail.com>
---
v2: Improve commit message

 meta/recipes-devtools/dnf/dnf_4.2.2.bb        | 2 ++
 meta/recipes-devtools/libdnf/libdnf_0.28.1.bb | 1 +
 2 files changed, 3 insertions(+)

diff --git a/meta/recipes-devtools/dnf/dnf_4.2.2.bb b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
index f38167f1ad..220f1aabbd 100644
--- a/meta/recipes-devtools/dnf/dnf_4.2.2.bb
+++ b/meta/recipes-devtools/dnf/dnf_4.2.2.bb
@@ -84,3 +84,5 @@ SYSTEMD_SERVICE_${PN} = "dnf-makecache.service dnf-makecache.timer \
                          dnf-automatic-notifyonly.service dnf-automatic-notifyonly.timer \
 "
 SYSTEMD_AUTO_ENABLE ?= "disable"
+
+PNBLACKLIST[dnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES', 'package_rpm', '', 'does not build correctly without package_rpm in PACKAGE_CLASSES', d)}"
diff --git a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
index 882c435b32..49afa04812 100644
--- a/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
+++ b/meta/recipes-devtools/libdnf/libdnf_0.28.1.bb
@@ -26,4 +26,5 @@ EXTRA_OECMAKE_append_class-native = " -DWITH_GIR=OFF"
 EXTRA_OECMAKE_append_class-nativesdk = " -DWITH_GIR=OFF"
 
 BBCLASSEXTEND = "native nativesdk"
+PNBLACKLIST[libdnf] ?= "${@bb.utils.contains('PACKAGE_CLASSES', 'package_rpm', '', 'does not build correctly without package_rpm in PACKAGE_CLASSES', d)}"
 
-- 
2.25.1