FESCo Agreed to Our Proposal

[Jeff Backus]

Hi all,

Good news! FESCo accepted our proposal and is giving us a chance to keep x86 support in Fedora! This is a temporary reprieve contingent upon our staying active and keeping up with the support issues.

FESCo did add two caveats to our proposal, which are:
* If the kernel team experiences build issues, i686 will be ExcludeArched with a block to FE-ExcludeArch-x86 until we can address them.
* FESCo would like to see details on supported hardware soon.

I haven't made it all the way through the logs yet, but I encourage the curious to read more here:
http://meetbot-raw.fedoraproject.org/teams/fesco/fesco.2017-09-08-16.00.log....

Special thanks to alexpl for attending the meeting.

In addition to the FE-ExcludeArch-x86 bug, I have created a general-purpose tracker bug for our purposes:
https://bugzilla.redhat.com/show_bug.cgi?id=1489998

I haven't figured out how to add mailing lists to the CC list yet. So until then, I encourage you to add yourself to the CC list.

I'll put together an e-mail to touch base with the kernel team.

jeff

[Alexander Ploumistos]

On Sat, Sep 9, 2017 at 5:02 AM, Jeff Backus  wrote:
> FESCo did add two caveats to our proposal, which are:
> If the kernel team experiences build issues, i686 will be ExcludeArched
>
> with a block to FE-ExcludeArch-x86 until we can address them.

Justin Forbes mentioned a build failure currently in rawhide (as far
as I can tell there hasn't been a bug report on it yet). Googling for
the error message (ERROR: "__udivdi3" [net/netfilter/xt_hashlimit.ko]
undefined!) I got a single result:
https://patchwork.ozlabs.org/patch/674586/
Any ideas on that one?
> FESCo would like to see details on supported hardware soon.

Should we start putting together a list of hardware available for
testing? If we don't have access to CPU X (and spare time to run
tests), we can't realistically hope to support it.
> I haven't made it all the way through the logs yet, but I encourage the
> curious to read more here:
> http://meetbot-raw.fedoraproject.org/teams/fesco/fesco.2017-09-08-16.00.log....
>
> Special thanks to alexpl for attending the meeting.

Athos was there too and his contribution was more substantial; I just
kept panicking over every proposal that fell on the table, e.g.
release validation testing…
> In addition to the FE-ExcludeArch-x86 bug, I have created a general-purpose
> tracker bug for our purposes:
> https://bugzilla.redhat.com/show_bug.cgi?id=1489998

We should try to make packagers aware of them both.

Have a nice weekend
Alex

[Jeff Backus]

On Sat, Sep 9, 2017 at 8:06 AM, Alexander Ploumistos
 wrote:
> On Sat, Sep 9, 2017 at 5:02 AM, Jeff Backus  wrote:
> > FESCo did add two caveats to our proposal, which are:
> > If the kernel team experiences build issues, i686 will be ExcludeArched
> >
> > with a block to FE-ExcludeArch-x86 until we can address them.
>
> Justin Forbes mentioned a build failure currently in rawhide (as far
> as I can tell there hasn't been a bug report on it yet). Googling for
> the error message (ERROR: "__udivdi3" [net/netfilter/xt_hashlimit.ko]
> undefined!) I got a single result:
> https://patchwork.ozlabs.org/patch/674586/
> Any ideas on that one?

Hmmm... Looks like it is related to doing 64bit divide on a 32bit
platform. According to a comment in this StackExchange thread, we
should be able to use the -g flag with GCC and then do an objdump -S
to disassemble the binary. We can then grep around.

Any update on this?
> > FESCo would like to see details on supported hardware soon.
>
> Should we start putting together a list of hardware available for
> testing? If we don't have access to CPU X (and spare time to run
> tests), we can't realistically hope to support it.

Yes, I think that is a good idea. I'll start another thread.
> > I haven't made it all the way through the logs yet, but I encourage the
> > curious to read more here:
> > http://meetbot-raw.fedoraproject.org/teams/fesco/fesco.2017-09-08-16.00.log....
> >
> > Special thanks to alexpl for attending the meeting.
>
> Athos was there too and his contribution was more substantial; I just
> kept panicking over every proposal that fell on the table, e.g.
> release validation testing…

Athos: My apologies - I still haven't finished reading the log. :)
Thank you for also being there.
> > In addition to the FE-ExcludeArch-x86 bug, I have created a general-purpose
> > tracker bug for our purposes:
> > https://bugzilla.redhat.com/show_bug.cgi?id=1489998
>
> We should try to make packagers aware of them both.

Agreed. I suppose we should send an e-mail to the packagers mailing
list. Any other suggestions?

jeff

[Alexander Ploumistos]

On Mon, Sep 11, 2017 at 7:52 PM, Jeff Backus  wrote:
> On Sat, Sep 9, 2017 at 8:06 AM, Alexander Ploumistos
>  wrote:
> > Justin Forbes mentioned a build failure currently in rawhide (as far
> > as I can tell there hasn't been a bug report on it yet). Googling for
> > the error message (ERROR: "__udivdi3" [net/netfilter/xt_hashlimit.ko]
> > undefined!) I got a single result:
> > https://patchwork.ozlabs.org/patch/674586/
> > Any ideas on that one?
>
> Hmmm... Looks like it is related to doing 64bit divide on a 32bit
> platform. According to a comment in this StackExchange thread, we
> should be able to use the -g flag with GCC and then do an objdump -S
> to disassemble the binary. We can then grep around.
>
> Any update on this?

Apparently this issue will resolve itself and it turns out there
wasn't much to do anyway:
https://bugzilla.redhat.com/show_bug.cgi?id=1490123

[Jeff Backus]

On Mon, Sep 11, 2017 at 1:04 PM, Alexander Ploumistos
 wrote:
> On Mon, Sep 11, 2017 at 7:52 PM, Jeff Backus  wrote:
> > On Sat, Sep 9, 2017 at 8:06 AM, Alexander Ploumistos
> >  wrote:
> > > Justin Forbes mentioned a build failure currently in rawhide (as far
> > > as I can tell there hasn't been a bug report on it yet). Googling for
> > > the error message (ERROR: "__udivdi3" [net/netfilter/xt_hashlimit.ko]
> > > undefined!) I got a single result:
> > > https://patchwork.ozlabs.org/patch/674586/
> > > Any ideas on that one?
> >
> > Hmmm... Looks like it is related to doing 64bit divide on a 32bit
> > platform. According to a comment in this StackExchange thread, we
> > should be able to use the -g flag with GCC and then do an objdump -S
> > to disassemble the binary. We can then grep around.
> >
> > Any update on this?
>
> Apparently this issue will resolve itself and it turns out there
> wasn't much to do anyway:
> https://bugzilla.redhat.com/show_bug.cgi?id=1490123

Great! Progress! :-D