[Mageia-dev] qemu new upstream release (1.0-rc1) and should we move from qemu-kvm to qemu?

[Michael Scherer]

Le dimanche 13 novembre 2011 à 22:32 +0100, Kamil Rytarowski a écrit :
> On 13.11.2011 10:58, Michael Scherer wrote:
> > Le samedi 12 novembre 2011 à 21:11 +0100, Kamil Rytarowski a écrit :
> >> On 12.11.2011 20:20, Michael Scherer wrote:
> >>> Le samedi 12 novembre 2011 à 16:44 +0100, Kamil Rytarowski a écrit :
> >>>
> >>>> There is also one important patch missed in Mageia -
> >>>> http://lists.gnu.org/archive/html/qemu-devel/2011-11/msg00787.html it's
> >>>> dependency for the GNS3 simulator. OpenSUSE already includes it
> >>>> https://build.opensuse.org/package/files?package=qemu&project=openSUSE%3ATools
> >>>>
> >>>> If nobody is against I will do it and contact the maintainer (misc).
> >>> I prefer to wait on the stable release ( ie, no rc1 ).
> >>> We will wait on stable version of qemu.
> >> OK
> >>> And no patch unless it comes from upstream ( and even, I am not keen on
> >>> backporting feature, better wait for stable release ).
> >>>
> >> GNS3 is already in stable! This package is broken - no dynamips (=no
> >> router emulation at all...), no patched qemu (no virtualization support
> >> at all...) According to the developers and their online documentation
> >> for package maintainers http://forum.gns3.net/post11571.html UDP patched
> >> Qemu is dependency/very important.
> > The fact that someone pushed a broken package is not a good reason to
> > add patches to qemu.
> OK, but I don't understand this.
> 
> Why to keep defunct packages (this could be at least "major+ issue"  on 
> our bugzilla) in stable and don't even want to fix, ignore this academic 
> software (with maybe overall 1 000 000* downloads and 100 000 regular 
> users), and to support... the inadvisable opinion of Mageia around.. at 
> least the GNS3 users.

Let me rephrase again. Everybody sooner or later think "that soft is
great, but why do not add just a small patch there". That's just one
patch, where is the problem ? 

The problem appear just after a few months, when the patch is still not
upstream, and that someone who do not know C, python whatever has to
take the software and maintain it. Or when someone who know how to
program lose time rediffing the patch instead of doing something more
useful. We face bugs that cannot be reproduced upstream, security
problem that could be added in non reviewed patch by devs. Fragmentation
in linux distributions are also caused by differents features, due to
patchs.

All of this need to be avoided, and I think we have enough problems with
stuff that people do not want to take care of it to not add more burden,
be it under the form of a small patch. All big collections start by one
little stuff. 


> * 799 968 Windows Downloads (just from the sourceforge mirrors) of the 
> latest Windows binary of GNS3 (source 
> http://sourceforge.net/projects/gns-3/files/GNS3/0.7.4/)
> 
> > We have too many patches on a general scale, and I
> > do not want to end with a 2nd package like gdb.
> >
> > Patches make harder to upgrade, harder to make sure security is done
> > correctly, and harder to ensure stuff are working ( since we are on our
> > own when we patch something ).
> > So for the patches, make sure it is upstream
> It's not qemu upstream, it's GNS3 and its community upstream.

If you want to have a feature in qemu, the road is "push it upstream".
Once accepted upstream, it will sooner or later be in our packages.

> >   ( and given the discussion
> > on ml, it should be soon )
> When I ask the developers, they don't know if qemu will include the 
> patch at all and when (now or after one year) and they suggested to do 
> the openSUSE way (today the most recommended and full featured Linux 
> distro for GNS3).

Maybe we are not talking of the same patch, but I am talking of this
one :
http://lists.gnu.org/archive/html/qemu-devel/2011-11/msg00629.html

AFAIK, the patch have been accepted, just not committed yet. The last
mail were from 1 week ago. The only issue is that they are in freeze for
now, and the git web interface is down, and I do see the commit in my
checkout about it so far.

> > and then in a tarball ( again, given that's a
> > rc 1, that should be ok soon ).
> >
> >> We must fix the package and provide at least not so heavy broken ones...
> >>
> >> I've prepared new version of GNS3, included into svn dynamips and
> >> xdotool (this one suggested) - these I can maintain with my mentor, so I
> >> ask for patch qemu in stable versus UDP support.
> > Updates are not supposed to get new features,
> Well this is a special case - the bugfix provides the feature, or the 
> feature provides the bugfix.

People will always tell "it is a special case". We can always say that
any feature is a bugfix, provided we say that the bug is "I cannot do
that". 


-- 
Michael Scherer