[kforge-dev] proposal: use buildbot and provide to test KForge [WAS: Re: trac tests - what a mess!]

[Rufus Pollock]

John Bywater wrote:
> Rufus Pollock wrote:
>> John Bywater wrote:
>>> Hi Rufus,
>>>
>>> I wasted several hours this afternoon with KForge. I'm totally
>>
>> I'm really sorry to hear that John.
> 
> Thanks. The good news (!) is that I've now got a tentative theory that 
> explains exactly why broken tests constitutes a problem. It also 
> addresses the "writing more than you thought you were going to write on 
> a wiki" phenomena you mentioned wanting to understand.
> 
> I'll try to run it past you when I see you next. ;-)

Great. I'll try calling you to schedule a meeting.

>>
>>> unimpressed that these tests don't actually work. I blame the crap
>>> testing regime I've been forced to accept. Any idea what the following
>>
>> I hope you haven't been 'forced' to accept any testing regime. I think 
>> we have both agreed that we would like to have more testing ...
> 
> I do feel unhappy that we're in a situation where some tests aren't run, 
> where we have things that aren't under test, where we don't get 
> immediate feedback when tests aren't passing, that sort of thing.

Well the buildbot is running on the main machine at present and it does 
report broken tests! For example it caught one recently which I detail 
in a separate mail.

> This wasn't my intention! So there are forces that have forced me into a 
> situation where, rather than a few minutes, I spend a few hours and 
> don't find my bug! As Data might say, this isn't within normal operating 
> parameters. :-)

Yes, though here it seems to be something with trac. However I agree 
with you that we have to be careful that we are not overwhelmed by the 
complexity and bugs are caught early ....

>>> is all about? I'm running Debian stable.
>>
>> Looking at the below this seems to be an issue in. trac/libsvn and not 
>> to do with kforge at all. I would note that when I upgraded from old 
>> debian stable to the new one I encountered problems with libsvn due to 
>> the fact related to the change from python2.3 to python2.4. It might 
>> be worth uninstalled (and purging) your python-subversion package and 
>> then reinstalling ...
> 
> Yes, I tried that, but still had the same. I agree, it probably isn't 
> KForge. I suspect the Trac version that comes with Debian.

I've tested locally now and can't reproduce that error so does look like 
it trac/svn related ...

> I propose we build a Buildbot-invoked Provide-managed KForge "daily 
> build and smoke" arrangement, which reports to the kforge-dev list.

I'm currently getting daily run deploy and run reports though it doesn't 
use provide but our own handrolled thing (kforge-autodeploy).

>>> Bored, bored, bored.....
>>
>> I too have spent hours tracking down more or less subtle bugs so you 
>> have my complete sympathies.
> 
> Thank you, that's appreciated. I know you have. Let's try to sort out 
> the testing regime? Perhaps we can now unify our efforts in respect of 
> the testing regime?

Yes, agree 100%.

~rufus