Re: i686 as secondary arch?

[Peter Robinson]

> On Tue, Jul 5, 2016 at 9:57 AM, Richard W.M. Jones 
> wrote:
>>
>> Timely article in the Register today:
>>
>> http://www.theregister.co.uk/2016/07/05/linux_letting_go_32bit_builds_on_the_way_out/
>>
>> I've been thinking about this as i686 is so often broken that I've now
>> stopped bothering to test it in the libguestfs tests that I do on
>> Rawhide:
>>
>>
>> http://pkgs.fedoraproject.org/cgit/rpms/libguestfs.git/commit/?id=aa63cef2d7679e1906551ef4e46c2e9a8861b56c
>>
>> If you need to run an i686 virtual machine based on Rawhide, my
>> experience is that it's more likely than not that it won't boot, and
>> no one cares.
>>
>> Do we have stats for the relative proportion of i686 vs x86-64
>> downloads?
>
>
> No really because of mirrors etc, but mirror manager stats from Feb
> (FPL DevConf talk) list i686 as around 20% unique IP hits, that
> doesn't take into account proxies/NAT using same IP etc.


 What clients are requesting from MirrorManager can also be seen here:

 https://admin.fedoraproject.org/mirrormanager/statistics/2016-07-05/archs
>>>
>>>
>>> These statistics do not cover package downloads of i686 packages which are
>>> part of the x86_64 repositories, do they?
>>>
>>> I think the numbers are also skewed by the fact that EPEL 7 is not available
>>> for i686, which is not of direct relevance to Fedora.  (The reason why it's
>>> missing is not lack of demand, but lack of a publicly available build root
>>> for Red Hat Enterprise Linux 7 on i686.)
>>>
>>
>> Here is a graph for just Fedora OS from time immemorial of Fedora
>> using a 7 day moving average.
>>
>> https://smooge.fedorapeople.org/simple_stats/fedora-hardware-full-ma.png
>>
>> I hope this is helpful.. [I am working on ways to make this available
>> regularly but am up to my neck in spam accounts so don;'t expect
>> soon.]
>
> That doesn't seem to distinguish between "ppc" and "ppc64" at first
> glance, which I would think we'd want it to.  Particularly since "ppc"
> isn't a thing that has existed in installable form for a while now.

We haven't built ppc (ie the 32 bit variant) since Fedora 20 so for
the recent history of Fedora with editions it's also irrelevant.

> Further, it doesn't distinguish between "ppc64" and "ppc64p7", which
> are actually separate architectures in the koji sense.

Yes, but they're not distributed as a separate repo over all and it's
a handful of packages that are built as ppc64p7 (currently 19) and
we're working to kill it off entirely.

> I realize the lines might be virtually invisible on the graph with a
> further breakdown, but I'm curious if the statistics themselves are
> being gathered properly there.
>
> josh
> --
> devel mailing list
> devel@lists.fedoraproject.org
> https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Josh Boyer]

On Tue, Jul 5, 2016 at 9:45 AM, Stephen John Smoogen  wrote:
> On 5 July 2016 at 06:46, Florian Weimer  wrote:
>> On 07/05/2016 11:09 AM, Adrian Reber wrote:
>>>
>>> On Tue, Jul 05, 2016 at 10:04:03AM +0100, Peter Robinson wrote:

 On Tue, Jul 5, 2016 at 9:57 AM, Richard W.M. Jones 
 wrote:
>
> Timely article in the Register today:
>
> http://www.theregister.co.uk/2016/07/05/linux_letting_go_32bit_builds_on_the_way_out/
>
> I've been thinking about this as i686 is so often broken that I've now
> stopped bothering to test it in the libguestfs tests that I do on
> Rawhide:
>
>
> http://pkgs.fedoraproject.org/cgit/rpms/libguestfs.git/commit/?id=aa63cef2d7679e1906551ef4e46c2e9a8861b56c
>
> If you need to run an i686 virtual machine based on Rawhide, my
> experience is that it's more likely than not that it won't boot, and
> no one cares.
>
> Do we have stats for the relative proportion of i686 vs x86-64
> downloads?


 No really because of mirrors etc, but mirror manager stats from Feb
 (FPL DevConf talk) list i686 as around 20% unique IP hits, that
 doesn't take into account proxies/NAT using same IP etc.
>>>
>>>
>>> What clients are requesting from MirrorManager can also be seen here:
>>>
>>> https://admin.fedoraproject.org/mirrormanager/statistics/2016-07-05/archs
>>
>>
>> These statistics do not cover package downloads of i686 packages which are
>> part of the x86_64 repositories, do they?
>>
>> I think the numbers are also skewed by the fact that EPEL 7 is not available
>> for i686, which is not of direct relevance to Fedora.  (The reason why it's
>> missing is not lack of demand, but lack of a publicly available build root
>> for Red Hat Enterprise Linux 7 on i686.)
>>
>
> Here is a graph for just Fedora OS from time immemorial of Fedora
> using a 7 day moving average.
>
> https://smooge.fedorapeople.org/simple_stats/fedora-hardware-full-ma.png
>
> I hope this is helpful.. [I am working on ways to make this available
> regularly but am up to my neck in spam accounts so don;'t expect
> soon.]

That doesn't seem to distinguish between "ppc" and "ppc64" at first
glance, which I would think we'd want it to.  Particularly since "ppc"
isn't a thing that has existed in installable form for a while now.
Further, it doesn't distinguish between "ppc64" and "ppc64p7", which
are actually separate architectures in the koji sense.

I realize the lines might be virtually invisible on the graph with a
further breakdown, but I'm curious if the statistics themselves are
being gathered properly there.

josh
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Florian Weimer]

On 07/05/2016 03:36 PM, Richard W.M. Jones wrote:

On Tue, Jul 05, 2016 at 12:56:37PM +0200, Florian Weimer wrote:

On 07/05/2016 10:57 AM, Richard W.M. Jones wrote:


If you need to run an i686 virtual machine based on Rawhide, my
experience is that it's more likely than not that it won't boot, and
no one cares.


Well, that's independent for the state as primary vs secondary architecture.

If we remove i686 as a primary architecture, we will not have i686
packages in the x86_64 repository.  Is this what we want?


IMHO that would be bad.


Okay, so we should keep building i686 within the same Koji instance 
(otherwise we'll have a lot of unnecessary pain).  For me, this is the 
dominant aspect of what constitutes a primary architecture.


Whether we need to provide installation images is a different question.

Florian
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Stephen John Smoogen]

On 5 July 2016 at 06:46, Florian Weimer  wrote:
> On 07/05/2016 11:09 AM, Adrian Reber wrote:
>>
>> On Tue, Jul 05, 2016 at 10:04:03AM +0100, Peter Robinson wrote:
>>>
>>> On Tue, Jul 5, 2016 at 9:57 AM, Richard W.M. Jones 
>>> wrote:

 Timely article in the Register today:

 http://www.theregister.co.uk/2016/07/05/linux_letting_go_32bit_builds_on_the_way_out/

 I've been thinking about this as i686 is so often broken that I've now
 stopped bothering to test it in the libguestfs tests that I do on
 Rawhide:


 http://pkgs.fedoraproject.org/cgit/rpms/libguestfs.git/commit/?id=aa63cef2d7679e1906551ef4e46c2e9a8861b56c

 If you need to run an i686 virtual machine based on Rawhide, my
 experience is that it's more likely than not that it won't boot, and
 no one cares.

 Do we have stats for the relative proportion of i686 vs x86-64
 downloads?
>>>
>>>
>>> No really because of mirrors etc, but mirror manager stats from Feb
>>> (FPL DevConf talk) list i686 as around 20% unique IP hits, that
>>> doesn't take into account proxies/NAT using same IP etc.
>>
>>
>> What clients are requesting from MirrorManager can also be seen here:
>>
>> https://admin.fedoraproject.org/mirrormanager/statistics/2016-07-05/archs
>
>
> These statistics do not cover package downloads of i686 packages which are
> part of the x86_64 repositories, do they?
>
> I think the numbers are also skewed by the fact that EPEL 7 is not available
> for i686, which is not of direct relevance to Fedora.  (The reason why it's
> missing is not lack of demand, but lack of a publicly available build root
> for Red Hat Enterprise Linux 7 on i686.)
>

Here is a graph for just Fedora OS from time immemorial of Fedora
using a 7 day moving average.

https://smooge.fedorapeople.org/simple_stats/fedora-hardware-full-ma.png

I hope this is helpful.. [I am working on ways to make this available
regularly but am up to my neck in spam accounts so don;'t expect
soon.]

> Florian
>
> --
> devel mailing list
> devel@lists.fedoraproject.org
> https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org



-- 
Stephen J Smoogen.
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Richard W.M. Jones]

On Tue, Jul 05, 2016 at 12:56:37PM +0200, Florian Weimer wrote:
> On 07/05/2016 10:57 AM, Richard W.M. Jones wrote:
> 
> >If you need to run an i686 virtual machine based on Rawhide, my
> >experience is that it's more likely than not that it won't boot, and
> >no one cares.
> 
> Well, that's independent for the state as primary vs secondary architecture.
> 
> If we remove i686 as a primary architecture, we will not have i686
> packages in the x86_64 repository.  Is this what we want?

IMHO that would be bad.

> (For me, armhfp is more more of a pain point due to slow build times.)

Agreed, but OTOH armv7 has active users.  Are we still using the
Calxeda machines to do the builds?  I have migrated my armv7 workloads
to virtual machines running on aarch64 hardware, and it certainly
"feels" like it performs better.

I'll try to post some numbers to back this up in a little while ...

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-p2v converts physical machines to virtual machines.  Boot with a
live CD or over the network (PXE) and turn machines into KVM guests.
http://libguestfs.org/virt-v2v
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Daniel P. Berrange]

On Tue, Jul 05, 2016 at 12:10:03PM +0100, Peter Robinson wrote:
> On Tue, Jul 5, 2016 at 11:56 AM, Florian Weimer  wrote:
> > On 07/05/2016 10:57 AM, Richard W.M. Jones wrote:
> >
> >> If you need to run an i686 virtual machine based on Rawhide, my
> >> experience is that it's more likely than not that it won't boot, and
> >> no one cares.
> >
> >
> > Well, that's independent for the state as primary vs secondary architecture.
> >
> > If we remove i686 as a primary architecture, we will not have i686 packages
> > in the x86_64 repository.  Is this what we want?
> 
> We're in the process of redefining what constitutes a secondary arch
> and this is part of that consideration. There's a bunch of proprietary
> common third party tools/apps that people rely on that still need i686
> around.

wine on x86_64 pulls in the i686 packages too, so it is not just closed
source stuff requiring i686. I don't see it being viable to drop support
for i686 on x86_64 any time soon.


Regards,
Daniel
-- 
|: http://berrange.com  -o-http://www.flickr.com/photos/dberrange/ :|
|: http://libvirt.org  -o- http://virt-manager.org :|
|: http://autobuild.org   -o- http://search.cpan.org/~danberr/ :|
|: http://entangle-photo.org   -o-   http://live.gnome.org/gtk-vnc :|
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Peter Robinson]

On Tue, Jul 5, 2016 at 11:56 AM, Florian Weimer  wrote:
> On 07/05/2016 10:57 AM, Richard W.M. Jones wrote:
>
>> If you need to run an i686 virtual machine based on Rawhide, my
>> experience is that it's more likely than not that it won't boot, and
>> no one cares.
>
>
> Well, that's independent for the state as primary vs secondary architecture.
>
> If we remove i686 as a primary architecture, we will not have i686 packages
> in the x86_64 repository.  Is this what we want?

We're in the process of redefining what constitutes a secondary arch
and this is part of that consideration. There's a bunch of proprietary
common third party tools/apps that people rely on that still need i686
around.

> (For me, armhfp is more more of a pain point due to slow build times.)

That will be changing soon, we have the replacement hardware and now
F-24 is out it's one of my primary focuses to get it into production.

Peter
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Adrian Reber]

On Tue, Jul 05, 2016 at 12:46:00PM +0200, Florian Weimer wrote:
> On 07/05/2016 11:09 AM, Adrian Reber wrote:
> > On Tue, Jul 05, 2016 at 10:04:03AM +0100, Peter Robinson wrote:
> > > On Tue, Jul 5, 2016 at 9:57 AM, Richard W.M. Jones  
> > > wrote:
> > > > Timely article in the Register today:
> > > > http://www.theregister.co.uk/2016/07/05/linux_letting_go_32bit_builds_on_the_way_out/
> > > > 
> > > > I've been thinking about this as i686 is so often broken that I've now
> > > > stopped bothering to test it in the libguestfs tests that I do on
> > > > Rawhide:
> > > > 
> > > > http://pkgs.fedoraproject.org/cgit/rpms/libguestfs.git/commit/?id=aa63cef2d7679e1906551ef4e46c2e9a8861b56c
> > > > 
> > > > If you need to run an i686 virtual machine based on Rawhide, my
> > > > experience is that it's more likely than not that it won't boot, and
> > > > no one cares.
> > > > 
> > > > Do we have stats for the relative proportion of i686 vs x86-64 
> > > > downloads?
> > > 
> > > No really because of mirrors etc, but mirror manager stats from Feb
> > > (FPL DevConf talk) list i686 as around 20% unique IP hits, that
> > > doesn't take into account proxies/NAT using same IP etc.
> > 
> > What clients are requesting from MirrorManager can also be seen here:
> > 
> > https://admin.fedoraproject.org/mirrormanager/statistics/2016-07-05/archs
> 
> These statistics do not cover package downloads of i686 packages which are
> part of the x86_64 repositories, do they?

No, that is not included. This is only what clients are sending as arch
in the mirrorlist/metalink request. Usually: arch=$basearch

Adrian
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Florian Weimer]

On 07/05/2016 10:57 AM, Richard W.M. Jones wrote:


If you need to run an i686 virtual machine based on Rawhide, my
experience is that it's more likely than not that it won't boot, and
no one cares.


Well, that's independent for the state as primary vs secondary architecture.

If we remove i686 as a primary architecture, we will not have i686 
packages in the x86_64 repository.  Is this what we want?


(For me, armhfp is more more of a pain point due to slow build times.)

Florian
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Florian Weimer]

On 07/05/2016 11:09 AM, Adrian Reber wrote:

On Tue, Jul 05, 2016 at 10:04:03AM +0100, Peter Robinson wrote:

On Tue, Jul 5, 2016 at 9:57 AM, Richard W.M. Jones  wrote:

Timely article in the Register today:
http://www.theregister.co.uk/2016/07/05/linux_letting_go_32bit_builds_on_the_way_out/

I've been thinking about this as i686 is so often broken that I've now
stopped bothering to test it in the libguestfs tests that I do on
Rawhide:

http://pkgs.fedoraproject.org/cgit/rpms/libguestfs.git/commit/?id=aa63cef2d7679e1906551ef4e46c2e9a8861b56c

If you need to run an i686 virtual machine based on Rawhide, my
experience is that it's more likely than not that it won't boot, and
no one cares.

Do we have stats for the relative proportion of i686 vs x86-64 downloads?


No really because of mirrors etc, but mirror manager stats from Feb
(FPL DevConf talk) list i686 as around 20% unique IP hits, that
doesn't take into account proxies/NAT using same IP etc.


What clients are requesting from MirrorManager can also be seen here:

https://admin.fedoraproject.org/mirrormanager/statistics/2016-07-05/archs


These statistics do not cover package downloads of i686 packages which 
are part of the x86_64 repositories, do they?


I think the numbers are also skewed by the fact that EPEL 7 is not 
available for i686, which is not of direct relevance to Fedora.  (The 
reason why it's missing is not lack of demand, but lack of a publicly 
available build root for Red Hat Enterprise Linux 7 on i686.)


Florian
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Nico Kadel-Garcia]

On Tue, Jul 5, 2016 at 6:02 AM, Richard W.M. Jones  wrote:
> On Tue, Jul 05, 2016 at 11:09:46AM +0200, Adrian Reber wrote:
>> On Tue, Jul 05, 2016 at 10:04:03AM +0100, Peter Robinson wrote:
>> > On Tue, Jul 5, 2016 at 9:57 AM, Richard W.M. Jones  
>> > wrote:
>> > > Timely article in the Register today:
>> > > http://www.theregister.co.uk/2016/07/05/linux_letting_go_32bit_builds_on_the_way_out/
>> > >
>> > > I've been thinking about this as i686 is so often broken that I've now
>> > > stopped bothering to test it in the libguestfs tests that I do on
>> > > Rawhide:
>> > >
>> > > http://pkgs.fedoraproject.org/cgit/rpms/libguestfs.git/commit/?id=aa63cef2d7679e1906551ef4e46c2e9a8861b56c
>> > >
>> > > If you need to run an i686 virtual machine based on Rawhide, my
>> > > experience is that it's more likely than not that it won't boot, and
>> > > no one cares.
>> > >
>> > > Do we have stats for the relative proportion of i686 vs x86-64 downloads?
>> >
>> > No really because of mirrors etc, but mirror manager stats from Feb
>> > (FPL DevConf talk) list i686 as around 20% unique IP hits, that
>> > doesn't take into account proxies/NAT using same IP etc.
>>
>> What clients are requesting from MirrorManager can also be seen here:
>>
>> https://admin.fedoraproject.org/mirrormanager/statistics/2016-07-05/archs
>
> More than I thought it would be.  I guess it wouldn't make sense to
> move i686 to a secondary arch while other secondary arches that might
> become primary (eg. aarch64) are still at far smaller numbers.
>
> Rich.

How much of that is for actual live systems, and how much of that is
for similarly obsolete build environments? I know that I wind up
testing my builds on both x86_64 and i386 mock environments for
testing, but I gave up on actually using the i386 installations years
ago.
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Richard W.M. Jones]

On Tue, Jul 05, 2016 at 11:09:46AM +0200, Adrian Reber wrote:
> On Tue, Jul 05, 2016 at 10:04:03AM +0100, Peter Robinson wrote:
> > On Tue, Jul 5, 2016 at 9:57 AM, Richard W.M. Jones  
> > wrote:
> > > Timely article in the Register today:
> > > http://www.theregister.co.uk/2016/07/05/linux_letting_go_32bit_builds_on_the_way_out/
> > >
> > > I've been thinking about this as i686 is so often broken that I've now
> > > stopped bothering to test it in the libguestfs tests that I do on
> > > Rawhide:
> > >
> > > http://pkgs.fedoraproject.org/cgit/rpms/libguestfs.git/commit/?id=aa63cef2d7679e1906551ef4e46c2e9a8861b56c
> > >
> > > If you need to run an i686 virtual machine based on Rawhide, my
> > > experience is that it's more likely than not that it won't boot, and
> > > no one cares.
> > >
> > > Do we have stats for the relative proportion of i686 vs x86-64 downloads?
> > 
> > No really because of mirrors etc, but mirror manager stats from Feb
> > (FPL DevConf talk) list i686 as around 20% unique IP hits, that
> > doesn't take into account proxies/NAT using same IP etc.
> 
> What clients are requesting from MirrorManager can also be seen here:
> 
> https://admin.fedoraproject.org/mirrormanager/statistics/2016-07-05/archs

More than I thought it would be.  I guess it wouldn't make sense to
move i686 to a secondary arch while other secondary arches that might
become primary (eg. aarch64) are still at far smaller numbers.

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-df lists disk usage of guests without needing to install any
software inside the virtual machine.  Supports Linux and Windows.
http://people.redhat.com/~rjones/virt-df/
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Adrian Reber]

On Tue, Jul 05, 2016 at 10:04:03AM +0100, Peter Robinson wrote:
> On Tue, Jul 5, 2016 at 9:57 AM, Richard W.M. Jones  wrote:
> > Timely article in the Register today:
> > http://www.theregister.co.uk/2016/07/05/linux_letting_go_32bit_builds_on_the_way_out/
> >
> > I've been thinking about this as i686 is so often broken that I've now
> > stopped bothering to test it in the libguestfs tests that I do on
> > Rawhide:
> >
> > http://pkgs.fedoraproject.org/cgit/rpms/libguestfs.git/commit/?id=aa63cef2d7679e1906551ef4e46c2e9a8861b56c
> >
> > If you need to run an i686 virtual machine based on Rawhide, my
> > experience is that it's more likely than not that it won't boot, and
> > no one cares.
> >
> > Do we have stats for the relative proportion of i686 vs x86-64 downloads?
> 
> No really because of mirrors etc, but mirror manager stats from Feb
> (FPL DevConf talk) list i686 as around 20% unique IP hits, that
> doesn't take into account proxies/NAT using same IP etc.

What clients are requesting from MirrorManager can also be seen here:

https://admin.fedoraproject.org/mirrormanager/statistics/2016-07-05/archs

Adrian
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org

Re: i686 as secondary arch?

[Peter Robinson]

On Tue, Jul 5, 2016 at 9:57 AM, Richard W.M. Jones  wrote:
> Timely article in the Register today:
> http://www.theregister.co.uk/2016/07/05/linux_letting_go_32bit_builds_on_the_way_out/
>
> I've been thinking about this as i686 is so often broken that I've now
> stopped bothering to test it in the libguestfs tests that I do on
> Rawhide:
>
> http://pkgs.fedoraproject.org/cgit/rpms/libguestfs.git/commit/?id=aa63cef2d7679e1906551ef4e46c2e9a8861b56c
>
> If you need to run an i686 virtual machine based on Rawhide, my
> experience is that it's more likely than not that it won't boot, and
> no one cares.
>
> Do we have stats for the relative proportion of i686 vs x86-64 downloads?

No really because of mirrors etc, but mirror manager stats from Feb
(FPL DevConf talk) list i686 as around 20% unique IP hits, that
doesn't take into account proxies/NAT using same IP etc.
--
devel mailing list
devel@lists.fedoraproject.org
https://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org