rpm: Recover RPM4 to OE-core

rpm: Recover RPM4 to OE-core

[Fan Xin]

Hi

I did some work to figure out whether the smartpm could been built with  
python3. However, I found smartpm is dependent with rpm, and OE-core  
uses RPM5 which seems has some bug.
http://rpm5.org/community/rpm-users/1098.html

RPM4 could be built with python3. So I think maybe we could recover RPM4  
into OE-core which is deleted
(oe-core a6e7a86f1635be9a688c56c25e9d215ea4d2cc84 )

The drop reason is that RPM4 is buggy and no one want to maintain it.

I build RPM4 in OE-core and the build is finished without any bug. And  
next I will try to use smartpm to manage RPM4 packages in my local  
environment.

By the way, could anyone give more detail info about the bugs in RPM4 or  
any suggestion about how to make smartpm built with python3.

Thanks.

Best Regards,

Fan

-- 

Re: rpm: Recover RPM4 to OE-core

[Joshua Lock]

Hi Fan,

On Thu, 2016-10-06 at 20:18 +0900, Fan Xin wrote:
> By the way, could anyone give more detail info about the bugs in RPM

We removed rpm4 early this year[1] because rather than have people be
surprised that it didn't work we decided to drop support for rpm4 until
such time as the project had sufficient resources to properly maintain
it.

There were at least two issues[2][3] reported that indicated rpm4
support was incomplete, though I believe they were mostly related to
Smart's interactions with RPM 4 rather than any issue in RPM 4 itself.

It's also worth pointing out that we're strongly considering dropping
SMART in the next (2.3) development cycle[4].

Do you have a reason for wanting rpm4 beyond the proven Python3
support?

Regards,

Joshua

1. http://lists.openembedded.org/pipermail/openembedded-core/2016-March
/118849.html
2. https://bugzilla.yoctoproject.org/show_bug.cgi?id=8968
3. https://bugzilla.yoctoproject.org/show_bug.cgi?id=8969
4. https://bugzilla.yoctoproject.org/show_bug.cgi?id=9675

Re: rpm: Recover RPM4 to OE-core

[Fan Xin]

Hi Joshua,

Thanks for your info.

Actually, our origin motivation is that we found smartpm is not 
maintained any more. We maintain smartpm for our distro at
https://github.com/ubinux/smart2

Then we realize smartpm should switch to python3. In order to solve this 
issue, there are four solutions as follows.

[1] smart/RPM5
[2] smart/RPM4
[3] dnf/RPM5
[4] dnf/RPM4

I confirm that RPM5 have some bugs with python3 and I am not familiar 
with RPM. So I think the easy way to solve this issue is [2], recover 
RPM4 and solve the bugs in smartpm.

>
> It's also worth pointing out that we're strongly considering dropping
> SMART in the next (2.3) development cycle[4].
>
Considering YP would like to drop smart in 2.3, I wonder YP would use 
[3]dnf/RPM5 or [4]dnf/RPM4 ?

> Do you have a reason for wanting rpm4 beyond the proven Python3
> support?
No, I just think python3 support issue should be solved. I am fine with 
RPM4 or RPM5 or smart or dnf.

Best Regards,

Fan

-- 

Re: rpm: Recover RPM4 to OE-core

[Maciej Borzęcki]

On Fri, Oct 7, 2016 at 5:19 AM, Fan Xin <fan.xin@jp.fujitsu.com> wrote:
> Hi Joshua,
>
> Thanks for your info.
>
> Actually, our origin motivation is that we found smartpm is not maintained
> any more. We maintain smartpm for our distro at
> https://github.com/ubinux/smart2
>
> Then we realize smartpm should switch to python3. In order to solve this
> issue, there are four solutions as follows.
>
> [1] smart/RPM5
> [2] smart/RPM4
> [3] dnf/RPM5
> [4] dnf/RPM4
>
> I confirm that RPM5 have some bugs with python3 and I am not familiar with
> RPM. So I think the easy way to solve this issue is [2], recover RPM4 and
> solve the bugs in smartpm.
>
>>
>> It's also worth pointing out that we're strongly considering dropping
>> SMART in the next (2.3) development cycle[4].
>>
> Considering YP would like to drop smart in 2.3, I wonder YP would use
> [3]dnf/RPM5 or [4]dnf/RPM4 ?

BTW. do you recall the rationale behind abandoning zypper in 1.4? I'm
curious about the motivation for using smart given its Python
dependency.

-- 
Maciej Borzecki
RnDity

Re: rpm: Recover RPM4 to OE-core

[Fan Xin]

Hi Maciej,

> BTW. do you recall the rationale behind abandoning zypper in 1.4? I'm
> curious about the motivation for using smart given its Python
> dependency.
>

The reason to drop Zypper is Zypper has too much components and brings 
huge potential support issues described in [1].

The maintainer of Yum is actively hostile toward RPM5. This is the 
reason why YP do not select YUM.

In [2], Smart is small and compact. And dependency of smart is small 
(python and rpm).

The dnf forks from Yum and dependent of hawkey, librepo, libcomps, 
libsolv. If according to the same raionale, smart is still the proper 
choice.

The relationship is complicated among rpm4, rpm5, dnf, yum, smart and 
OE-core. And thing is changing makes it harder to figure out this issue.

Best Regards,

Fan

[1] https://lists.yoctoproject.org/pipermail/yocto/2012-October/010384.html
[2] 
http://lists.openembedded.org/pipermail/openembedded-core/2015-January/100582.html



-- 

Re: rpm: Recover RPM4 to OE-core

[Maciej Borzęcki]

On Fri, Oct 7, 2016 at 11:52 AM, Fan Xin <fan.xin@jp.fujitsu.com> wrote:
> Hi Maciej,
>
>> BTW. do you recall the rationale behind abandoning zypper in 1.4? I'm
>> curious about the motivation for using smart given its Python
>> dependency.
>>
>
> The reason to drop Zypper is Zypper has too much components and brings huge
> potential support issues described in [1].
>
> The maintainer of Yum is actively hostile toward RPM5. This is the reason
> why YP do not select YUM.
>
> In [2], Smart is small and compact. And dependency of smart is small (python
> and rpm).
>
> The dnf forks from Yum and dependent of hawkey, librepo, libcomps, libsolv.
> If according to the same raionale, smart is still the proper choice.
>
> The relationship is complicated among rpm4, rpm5, dnf, yum, smart and
> OE-core. And thing is changing makes it harder to figure out this issue.
>
> Best Regards,
>
> Fan
>
> [1] https://lists.yoctoproject.org/pipermail/yocto/2012-October/010384.html
> [2]
> http://lists.openembedded.org/pipermail/openembedded-core/2015-January/100582.html
>
Thanks. This has been an interesting read.

-- 
Maciej Borzecki
RnDity

Re: rpm: Recover RPM4 to OE-core

[Richard Purdie]

On Fri, 2016-10-07 at 09:20 +0200, Maciej Borzęcki wrote:
> BTW. do you recall the rationale behind abandoning zypper in 1.4? I'm
> curious about the motivation for using smart given its Python
> dependency.

zypper didn't cross compile easily and had a fairly nasty dependency
chain if I remember rightly including things like boost. We had quite
an extensive patchset against it and libzypp which could never be
upstreamed:

http://git.yoctoproject.org/cgit.cgi/poky/commit/?id=fcdc8d7f6d10b3abf9
f6a28cff55b1cc89b88bc9

zypper was also hostile to rpm5 if I remember rightly.

Switching to something python based was much lower overhead on the
system compared to C++ and boost.

Cheers,

Richard

Re: rpm: Recover RPM4 to OE-core

[Joshua Lock]

On Fri, 2016-10-07 at 12:19 +0900, Fan Xin wrote:
> Hi Joshua,
> 
> Thanks for your info.
> 
> Actually, our origin motivation is that we found smartpm is not 
> maintained any more. We maintain smartpm for our distro at
> https://github.com/ubinux/smart2

Indeed, the lack of maintenance for Smart is one of the reasons we're
looking to switch away from it.

> Then we realize smartpm should switch to python3. In order to solve
> this 
> issue, there are four solutions as follows.
> 
> [1] smart/RPM5
> [2] smart/RPM4
> [3] dnf/RPM5
> [4] dnf/RPM4
> 
> I confirm that RPM5 have some bugs with python3 and I am not
> familiar 
> with RPM. So I think the easy way to solve this issue is [2],
> recover 
> RPM4 and solve the bugs in smartpm.
> 
> > 
> > 
> > It's also worth pointing out that we're strongly considering
> > dropping
> > SMART in the next (2.3) development cycle[4].
> > 
> Considering YP would like to drop smart in 2.3, I wonder YP would
> use 
> [3]dnf/RPM5 or [4]dnf/RPM4 ?

There are two main reasons we want to switch away from Smart. Firstly
we made a concerted effort to switch to Python3 in the 2.2 cycle —
Smart is one of the few recipes in OE-Core which still pulls in
Python2, maybe the only one we use by default (for RPM using distros).

Secondly Smart is unmaintained and has several known issues related to
its abstraction over package backends.

So far as I'm aware no decision has been made as to what we'll replace
Smart with yet. I believe the first step will be to investigate RPM5 &
dnf.

Regards,

Joshua