[oi-dev] Testing Updating packages and hipster-testing repo.

[Nikola M.]

On 10/ 9/13 09:05 AM, Alexander Pyhalov wrote:
> Hello.
>
> On 10/09/2013 10:55, Nikola M. wrote:
>
>> What do you think?
>>
>> Nikola M.
>>
>>
>
> I think this doesn't work - just not enough users even in /hipster. 
> Will someone ever use /hipster-testing?
>
> For example. I've just recompiled hpijs. But I personally don't use 
> it. How long will it take to find someone to test it? A week or more?
Putting anything in a public repo without some form of testing is not 
great practice
There must be someone using it & is subscribed to the list or other 
means of communication with the users
Like having user registration procedure upon install and automatic 
subscribing to some PR mailing lists might help with that. There are 
many benefits of being registered user , like being informed when 
packages you use have been changed and to take measure something you 
depend on does not break on update. That line of direct communication 
have benefits of including more people in the process on many levels.

We that are willing to test things should at least test it however we 
can. (In few days top)
But that publisher before 'hipster', could test upgrading and if 
packaging is right, all those distribution stuff that have nothing to do 
with functionality of the package itself etc.
And at least we could say "we have testing procedure before publishing it"
and quality control exists but needs to be improved".
>
> I've already mentioned - I think about /hipster as about 
> FreeBSD-current. Usually it works and everyone tries to avoid breaking 
> it. But sometimes it happens.
>
> I'd prefer the opposite approach - to  make /hipster-stable :) But who 
> and when will update it?
That is also OK, but there is another view - people get used that 
Hipster works.
If Hipster is broken for something for some time period (and there is no 
fix for it for some time) then it is a bad practice.

I think it is important to have Hipster better and avoid obvious 
breaking and obvious catastrophic changes of functionality by mandatory 
testing before putting it in hipster.

It is not good just to 'dump changes' to the hipster and hope for the 
best. Alt least one more stage is needed before that that is -testing- 
with at least one confirmation before putting it in hipster.

And hipster-stable is great idea,
I suppose hipster-testing could have a track on what is actually good to 
go there and what is not,
even if it lives in hipster. Hipster-stable could be points from where 
people that test things, could test upgrades and could upgrade to, 
before going further with hipster update.

So most of people will continue having just hipster installed. But we 
would have
hipster-testing before and hipster-stable after, to help testing crew.

After few hipster-stable , That could be good enough to go to 
Openindiana /dev (It is actually Openindiana release). All paying road 
to Openindiana /stable and Openindiana /updates ,
that would need another queue of testing and upgrading, but that is for 
future. (as well as having SFE also synchronized and not broken etc).

Thing is, people will test something that is most of the time usable.
And with extremely low or non-existent record of breakage.
In times with bigger audience (opensolaris/dev) it surely had bugs and 
so on, but it was usable and updated all the time, so people could use 
it and report bugs.
For OI, that is Hipster now. So testing is needed for usability.

N.M.