Fixing livecd ISO labels (bz 1145264)

[Cole Robinson]

Hi all,

The ISO volume ID for the Fedora 21 workstation livecd is 
'Fedora-Live-Workstation-x86_64-2', which doesn't actually say 21 in it, and 
libosinfo can't use it to detect the fedora version for boxes/virt-manager.

Koji is responsible for generating the volume ID, and there's a bug filed:

https://bugzilla.redhat.com/show_bug.cgi?id=1145264

I also sent a (untested) koji patch to buildsys list:

https://lists.fedoraproject.org/pipermail/buildsys/2014-October/004411.html

However neither have received a response, so I'm reaching out here.

If this isn't fixed before F21 GA, we are stuck with the crappy volume ID 
forever and tools that are depending on it are in a tough spot. Is there any 
process to escalate this and get the fix deployed to infrastructure? (FWIW 
I've just proposed the bug as a freeze exception, but presumably it will take 
more coordination that that)

Thanks,
Cole

[Cole Robinson]

Fixing buildsys CC

On 11/15/2014 01:29 PM, Cole Robinson wrote:
>  Hi all,
>
>  The ISO volume ID for the Fedora 21 workstation livecd is
>  'Fedora-Live-Workstation-x86_64-2', which doesn't actually say 21 in it, and
>  libosinfo can't use it to detect the fedora version for boxes/virt-manager.
>
>  Koji is responsible for generating the volume ID, and there's a bug filed:
>
>  https://bugzilla.redhat.com/show_bug.cgi?id=1145264
>
>  I also sent a (untested) koji patch to buildsys list:
>
>  https://lists.fedoraproject.org/pipermail/buildsys/2014-October/004411.html
>
>  However neither have received a response, so I'm reaching out here.
>
>  If this isn't fixed before F21 GA, we are stuck with the crappy volume ID
>  forever and tools that are depending on it are in a tough spot. Is there any
>  process to escalate this and get the fix deployed to infrastructure? (FWIW
>  I've just proposed the bug as a freeze exception, but presumably it will take
>  more coordination that that)
>
>  Thanks,
>  Cole

[Dennis Gilmore]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sat, 15 Nov 2014 13:31:42 -0500
Cole Robinson <crobinso(a)redhat.com&gt; wrote:

>  Fixing buildsys CC
>  
>  On 11/15/2014 01:29 PM, Cole Robinson wrote:
>  > Hi all,
>  >
>  > The ISO volume ID for the Fedora 21 workstation livecd is
>  > 'Fedora-Live-Workstation-x86_64-2', which doesn't actually say 21
>  > in it, and libosinfo can't use it to detect the fedora version for
>  > boxes/virt-manager.
>  >
>  > Koji is responsible for generating the volume ID, and there's a bug
>  > filed:
>  >
>  > https://bugzilla.redhat.com/show_bug.cgi?id=1145264
>  >
>  > I also sent a (untested) koji patch to buildsys list:
>  >
>  > https://lists.fedoraproject.org/pipermail/buildsys/2014-October/004411.html
>  
NAK on this patch its the completely wrong place to apply such a thing.
the fix belongs in livecd-tools

>  > However neither have received a response, so I'm reaching
> out here.
>  >
>  > If this isn't fixed before F21 GA, we are stuck with the crappy
>  > volume ID forever and tools that are depending on it are in a tough
>  > spot. Is there any process to escalate this and get the fix
>  > deployed to infrastructure? (FWIW I've just proposed the bug as a
>  > freeze exception, but presumably it will take more coordination
>  > that that)
>  
Dennis
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v2
>
> iQIcBAEBAgAGBQJUaUrOAAoJEH7ltONmPFDRgEEP/j+bPJnFWTwQRJ7oYOne9fAy
> pKWSO0YCcyA0FcXmIBO3ldUUsCUDCA1gM2paC20NGO63MFGV6ih0NkU6PBiIWeyS
> 31V89xt+7yuK8GR7Cm6G3Jeae1Esi25NapVsPmFnaP7+m4cs640n149S5jJtDORl
> lXtnJy6Ojfj+s/AEpjEHpLH+9E90woTvGQnNMPKE8IOv0Zy5VT9eigqmkoO8qMah
> cdB/MNzBKIvV9OqCqQgEotZnfs2RszqYw/uR1CyppASl1JPa0NlfBqE64zfbtPKZ
> FF0q9km7XYAqS2b5RUzuDxDE7smKozwDqroe3MH+hzbhUjvUkS1cPsG+8GJqLoCP
> nkty6HyF7SCWWYFnopfGUJjlTa4jxapxtai+uoQl+prfzF1iiRCiWHzgSFdS00Ps
> dhuULjBzwC3zW5UHF0PN7e7JT3wSmR8k+PBLtLo3qW5zBNuPocxkagDznrWdj5oj
> myQYx7g0Aiw5aHKz8V+AroBb1bT4BVz6clk/3oJz+8QhGVOwyKBoIXQ36ctbfalA
> ncM0sWKBVBYz8rZnBbLOO4RwyvKnsnTJQjvL+UNA5RAPAQzvIlYrkv0MnFgTU20R
> IYygBJmk5yPu5tgQOBlq9FIF6CdhLo715RCbfXS2cRVEk2QbICerlW/IkGLo6PBT
> a8DkBQtok2qAMvyrvT7F
> =N7wO
> -----END PGP SIGNATURE-----
>

[Dennis Gilmore]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sun, 16 Nov 2014 19:09:25 -0600
Dennis Gilmore <dennis(a)ausil.us&gt; wrote:

>  -----BEGIN PGP SIGNED MESSAGE-----
>  Hash: SHA1
>  
>  On Sat, 15 Nov 2014 13:31:42 -0500
>  Cole Robinson <crobinso(a)redhat.com&gt; wrote:
>  
>  > Fixing buildsys CC
>  > 
>  > On 11/15/2014 01:29 PM, Cole Robinson wrote:
>  > > Hi all,
>  > >
>  > > The ISO volume ID for the Fedora 21 workstation livecd is
>  > > 'Fedora-Live-Workstation-x86_64-2', which doesn't actually say 21
>  > > in it, and libosinfo can't use it to detect the fedora version for
>  > > boxes/virt-manager.
>  > >
>  > > Koji is responsible for generating the volume ID, and there's a
>  > > bug filed:
>  > >
>  > > https://bugzilla.redhat.com/show_bug.cgi?id=1145264
>  > >
>  > > I also sent a (untested) koji patch to buildsys list:
>  > >
>  > > https://lists.fedoraproject.org/pipermail/buildsys/2014-October/004411.html
>  
>  NAK on this patch its the completely wrong place to apply such a
>  thing. the fix belongs in livecd-tools
>  
of course though things need to be done differently in koji. because we
would need to pass through the full untrunckated value or do something
else that involves ignoring the CLI. But I do stand by koji being the
wrong place for that knowledge. especially true if we change the naming
scheme going forward. which is something we intend to do. keep them
under 32 characters but make them a bit more readable. really libosinfo
should be looking in the image and not relying on the filesystem
volumeid to determine the release of an image. something outside of the
contents could be manipulated, while it is much less likely to happen
inside the image.

Denni

>  > > However neither have received a response, so I'm
> reaching out
>  > > here.
>  > >
>  > > If this isn't fixed before F21 GA, we are stuck with the crappy
>  > > volume ID forever and tools that are depending on it are in a
>  > > tough spot. Is there any process to escalate this and get the fix
>  > > deployed to infrastructure? (FWIW I've just proposed the bug as a
>  > > freeze exception, but presumably it will take more coordination
>  > > that that)
>  
> >  Dennis
> >  
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v2
>
> iQIcBAEBAgAGBQJUaU6qAAoJEH7ltONmPFDRFTcP/3b1fbWwIkvrLMRfnwfDrfVG
> HvdiG8Jnk17mDD4GMcJBcdeXiRh1DPDlL2NUAcdrYIXCUpY9ObxCmDRgpqy8Q5mb
> KMoDXutfhM6qjyJfGrQ/lXuidL/13EwcnrRkTlJb/asD9MxjDbfE6WoDasUXGE+/
> lvBiiieY8n+n2bgz4Z8xq3Y3p0nPoq0m+GOiw+LbwE/8gLm9hxjifx6BMaFpFmoJ
> RSo+HT02xa9X6yu4c2uE//0OsPRZDXc9v2wAjf0gUqPuoHHG3aAC9AhO2DN1Ef74
> /eXQFA6FbmYIGLsvLrdh+IeEdtZw8I1gyIoR5OFC/BoNG/v9h+KUDi6Dk6+YqPU7
> tRIDE1HMTyEF0vOqL6Rj8QcdrECYflxt4P9HtD65xPZ2zYj3Zg5tMzcIDI4HNrOG
> 2tkumr51COsL/zimKySL1ETARXY0wIjExEnfIT9zPb1Olww3E8PhYotszLjsuYJH
> F8vdFZhea/UYLZsREMcI+gc157WsGCGK6juhQqMa8DnEvXkjT7sVkjWzssj/4GE/
> fS3tPPAuzLW/dcjP6fnIrJTpf/VoajcTeYqz7sH6AQmqo1taxOBxR+pQiwbrDHb6
> gpbjs9CdcfTt8LHfYk8JObCa8wRec0xH7NpWVAGgS67qtuYzcMyZ3uBzTMbHgqg0
> cL+JgMO8VWZG1VFxd1W7
> =4ACh
> -----END PGP SIGNATURE-----
>

[Cole Robinson]

On 11/16/2014 08:26 PM, Dennis Gilmore wrote:
>  -----BEGIN PGP SIGNED MESSAGE-----
>  Hash: SHA1
>
>  On Sun, 16 Nov 2014 19:09:25 -0600
>  Dennis Gilmore <dennis(a)ausil.us&gt; wrote:
>
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> >
> > On Sat, 15 Nov 2014 13:31:42 -0500
> > Cole Robinson <crobinso(a)redhat.com&gt; wrote:
> >
> >> Fixing buildsys CC
> >>
> >> On 11/15/2014 01:29 PM, Cole Robinson wrote:
> >>> Hi all,
> >>>
> >>> The ISO volume ID for the Fedora 21 workstation livecd is
> >>> 'Fedora-Live-Workstation-x86_64-2', which doesn't actually say
> 21
> >>> in it, and libosinfo can't use it to detect the fedora version for
> >>> boxes/virt-manager.
> >>>
> >>> Koji is responsible for generating the volume ID, and there's a
> >>> bug filed:
> >>>
> >>> https://bugzilla.redhat.com/show_bug.cgi?id=1145264
> >>>
> >>> I also sent a (untested) koji patch to buildsys list:
> >>>
> >>> https://lists.fedoraproject.org/pipermail/buildsys/2014-October/004411.html
> >
> > NAK on this patch its the completely wrong place to apply such a
> > thing. the fix belongs in livecd-tools
>
>  of course though things need to be done differently in koji. because we
>  would need to pass through the full untrunckated value or do something
>  else that involves ignoring the CLI. But I do stand by koji being the
>  wrong place for that knowledge. especially true if we change the naming
>  scheme going forward. which is something we intend to do. keep them
>  under 32 characters but make them a bit more readable. 
ccing bcl who originally moved the bug to koji. I'm all for long term fixup 
but I'd also really like to see some solution before F21 GA. Ideas anyone?

really libosinfo
>  should be looking in the image and not relying on the filesystem
>  volumeid to determine the release of an image. something outside of the
>  contents could be manipulated, while it is much less likely to happen
>  inside the image.
>  
Yeah that would be a useful capability to have, but it's a whole different 
world of hurt and a lot more work that isn't going to materialize anytime in 
the short to midterm. And honestly just depending on volume ID has worked out 
pretty well already.

- Cole

> >>> However neither have received a response, so I'm reaching out
> >>> here.
> >>>
> >>> If this isn't fixed before F21 GA, we are stuck with the crappy
> >>> volume ID forever and tools that are depending on it are in a
> >>> tough spot. Is there any process to escalate this and get the fix
> >>> deployed to infrastructure? (FWIW I've just proposed the bug as a
> >>> freeze exception, but presumably it will take more coordination
> >>> that that)
> >
> > Dennis

[Matthew Miller]

On Mon, Nov 17, 2014 at 04:14:11PM -0500, Cole Robinson wrote:
>  On 11/16/2014 08:26 PM, Dennis Gilmore wrote:
>  >>NAK on this patch its the completely wrong place to apply such a
>  >>thing. the fix belongs in livecd-tools 
[...]
>  ccing bcl who originally moved the bug to koji. I'm all for long
>  term fixup but I'd also really like to see some solution before F21
>  GA. Ideas anyone? 
Unless there is another easy change that can be made (suggestions,
Dennis?), I think we should Do Whatever Works in the short-term fix,
and then file a ticket for long-term fix, including reverting the
short-term one if necessary. I appreciate that this adds yet more cruft
to an already messy system, but in the absence of alternatives....


-- 
Matthew Miller
<mattdm(a)fedoraproject.org&gt;
Fedora Project Leader

[Brian C. Lane]

On Tue, Nov 18, 2014 at 07:47:19AM -0500, Matthew Miller wrote:
>  On Mon, Nov 17, 2014 at 04:14:11PM -0500, Cole Robinson wrote:
>  > On 11/16/2014 08:26 PM, Dennis Gilmore wrote:
>  > >>NAK on this patch its the completely wrong place to apply such a
>  > >>thing. the fix belongs in livecd-tools
>  [...]
>  > ccing bcl who originally moved the bug to koji. I'm all for long
>  > term fixup but I'd also really like to see some solution before F21
>  > GA. Ideas anyone?
>  
>  Unless there is another easy change that can be made (suggestions,
>  Dennis?), I think we should Do Whatever Works in the short-term fix,
>  and then file a ticket for long-term fix, including reverting the
>  short-term one if necessary. I appreciate that this adds yet more cruft
>  to an already messy system, but in the absence of alternatives.... 
This kind of information doesn't belong in livecd-creator. You can pass
it a volume id. Koji knows (or should know) what it is building, it
should construct the correct name and pass it. The default in
livecd-creator is just a default and is used to match up the isolinux
config with the image for booting, which is working just fine.

Also, can we stop with the quick fixes? They live forever, so let's do
this the right way.

-- 
Brian C. Lane | Anaconda Team | IRC: bcl #anaconda | Port Orchard, WA (PST8PDT)

[Matthew Miller]

On Tue, Nov 18, 2014 at 08:46:54AM -0800, Brian C. Lane wrote:
>  This kind of information doesn't belong in livecd-creator. You
> can pass
>  it a volume id. Koji knows (or should know) what it is building, it
>  should construct the correct name and pass it. The default in
>  livecd-creator is just a default and is used to match up the isolinux
>  config with the image for booting, which is working just fine.
>  
>  Also, can we stop with the quick fixes? They live forever, so let's do
>  this the right way. 
Agree in general, but a) sometimes, that's where we are, and b) as I
understand it, the F22 plan is to use imagefactory and/or
livemedia-creator rather than livecd creator, so it really is
short-lived.

-- 
Matthew Miller
<mattdm(a)fedoraproject.org&gt;
Fedora Project Leader

[Kevin Fenzi]

On Tue, 18 Nov 2014 12:52:16 -0500
Matthew Miller <mattdm(a)fedoraproject.org&gt; wrote:

>  On Tue, Nov 18, 2014 at 08:46:54AM -0800, Brian C. Lane wrote:
>  > This kind of information doesn't belong in livecd-creator. You can
>  > pass it a volume id. Koji knows (or should know) what it is
>  > building, it should construct the correct name and pass it. The
>  > default in livecd-creator is just a default and is used to match up
>  > the isolinux config with the image for booting, which is working
>  > just fine.
>  > 
>  > Also, can we stop with the quick fixes? They live forever, so let's
>  > do this the right way.
>  
>  Agree in general, but a) sometimes, that's where we are, and b) as I
>  understand it, the F22 plan is to use imagefactory and/or
>  livemedia-creator rather than livecd creator, so it really is
>  short-lived. 
well, it has to live for f21 respins basically until f21 goes eol. ;( 

From my update to the bug just now: 

"So we discussed this today in IRC.

All the choices are pretty bad, but the koji patch for now is the least
bad. ;( 

It does mean we will have to carry that patch forward until f21 is
eol... since we may have to do respins. ;( 

Dennis is going to make a patched koji package and put in for a infra
freeze break to apply it to builders."

kevin