Re: [Xen-users] xen forum

[Gordan Bobic <gordan@bobich.net>]

On 05/24/2013 10:36 PM, Stefano Stabellini wrote:

>>>>>> It would be easier for us if the bug reports and such were posted on
>>>>>> xen-devel.
>>>>>> Please consult http://www.chiark.greenend.org.uk/~sgtatham/bugs.html
>>>>>> when
>>>>>> doing it.
>>>>>>
>>>>>
>>>>>
>>>>> My own experience is that posts (at least from me) are regularly
>>>>> missed/ignored on the devel list, including a signed patch, so I
>>>>> personally think a bug tracker would be a better option. Bug trackers
>>>>> don't (or at least shouldn't :) ) forget or miss. That's they're raison
>>>>> d'etre. I honestly don't know how anyone can do business using this
>>>>> list, but that's just my humble opinion.
>>>>
>>>> Did you also look in the MAINTAINERS file to make sure you copied the
>>>> right
>>>> maintainer?
>>>>
>>>> The reason for skipping the Bugzilla system is that it is soo out of date
>>>> that
>>>> we don't use it anymore.
>>>
>>>
>>> Actually I recall there is a secondary reason too - which is that we get
>>> copied
>>> on distros bugs that affect Xen. For example in Fedora I (and Dariof) get
>>> copied on
>>> any Linux kernel issues that are related to Xen. In Debian I believe Ian
>>> Campbell
>>> gets copied as well. For SuSE it is Jan and Olaf. Not sure about the other
>>> distros.
>>>
>>> And then if you use Oracle Linux, I get copied too. Then there is the
>>> internal bug system
>>> if you using OVM and the Linux kernel bug-system where I get copied too.
>>>
>>> That is a lot of bug systems to keep track of - and since most of the users
>>> use a
>>> distro they end up using their distro bug-system. And then Xen's bugzilla
>>> system
>>> became less and less important to keep track of stuff.
>>>
>>> Oh, and there are the five mailing lists and the fire-hose LKML. Yuck, soo
>>> many emails.
>>
>> Surely the sensible thing to do is to have one Xen bug tracking system and
>> only use that. If distro maintainers wish to file bugs in the Xen bug tracker
>> for Xen bugs, they are free to do so, same as any other user. Xen is the
>> upstream project - Xen bugs should be fed from distros up to Xen, not the
>> other way around. Xen bugs are then tracked with the single Xen bug tracker
>> and they are all in one place, searchable reviewable and easy to keep track
>> of. Is this not obvious? Am I missing missing an issue that has been
>> too-subtly implied but not explicitly stated?
>
> In an ideal world maybe. What usually happens is that distros keep using
> their bug trackers and keep recommending their users to fill bug to
> them. These bug trackers get out of sync with the upstream bug tracker.

That's distro problem, not a Xen problem, and should not be expected to 
be a Xen problem, nor should it ever become a Xen problem.

> Moreover some people don't use bug trackers and submit bugs as emails
> anyway, as a consequence the bug tracker usually needs to be kept
> up-to-date manually by one or more members of the community.

So stop accepting emailed bug reports. If somebody emails one, tell them 
to create a bugzilla account and file it there. If that is too hard, 
they clearly don't care about the bug enough. It's no better or worse a 
filtering system than seeing who is going to bother bumping an email 
thread if it gets missed.

> In the long run they tend to be "left behind".
> In Linux it has been tried several times to introduce bug trackers,
> most of the times failing completely.

My preferred Linux distribution uses a bugzilla bug tracker and it works 
very well indeed.

>>> It is mostly FIFO with the 'oh wow, this needs to be fixed NOW!' preempting
>>> it.
>>> In all honestly it sucks as a track system, but I am not really sure of how
>>> else to do this
>>> without spending a massive time doing 'click here on this button and add
>>> this comment,
>>> set dependency on this bug' and instead concentrate my time in an editor.
>>>
>>> I believe we need something that can bridge both of these - helping
>>> developers to
>>> know about bugs and also track them so users know that things are done and
>>> not ignored.
>>> And so low maintaince for developers that they can focus on looking at code
>>> all day.
>>
>> I don't think this is a new problem, and I do think the problem has been
>> solved many times and solved well. If there is an obvious flaw in what I said
>> above, please do point it out. But claiming that a broadcast system is bad and
>> therefore ignoring a single-point tracking system is the way forward is as
>> much of a contradiction in terms as I can imagine on this subject.
>
> It is not a new problem but it has never been solved properly, just give
> a look at the status of bug trackers in the linux kernel to get an idea.
> Lunchpad was supposed to be the bug tracker to rule them all, but it
> ended up being just one more bug tracker.

It works just fine for RH and Fedora. So clearly the problem must be in 
something else.

> That said, I don't mean that it's all hopeless and doomed, you certainly
> raised some good points and I think we have room for improvement.
> It's just not as simple as it seems.
> Personally I am in favor of introducing a bug tracker if we have a way
> to integrate it into our current process and make sure it's kept up to
> date.

I sincerely hope it happens.

Gordan