[Buildbot-devel] p4poller.py changes detection since last sync

[Yasser LEBON]

Of course.

Yasser


2013/10/28 Bill Deegan <bill at baddogconsulting.com>

> Yasser,
>
> Excellent. Can you submit a pull request with your changes?
>
> -Bill
>
>
> On Mon, Oct 28, 2013 at 12:48 PM, Yasser LEBON <yasser.lebon at gmail.com>wrote:
>
>> I totally agree with you that the response is too vague. My impression is
>> that they don't know exactly where the performance differences comes from
>> but they suspects the db queries .
>>
>> In any case, using the head changelist number instead of @NOW improves
>> definitively the hit on our server.
>>
>> Yasser
>>
>>
>> 2013/10/28 William Deegan <bill at baddogconsulting.com>
>>
>>> Yasser,
>>>
>>>
>>> On Oct 28, 2013, at 12:03 PM, Yasser LEBON <yasser.lebon at gmail.com>
>>> wrote:
>>>
>>> After clarification with team responsible of Perforce in my company it
>>> seems that using @HEAD or @NOW the scanning occurs at the level of the
>>> db.rev instead of db.change that gives differents impacts on Perforce
>>> server. I contacted Perforce and here is their answer:
>>> "The difference has to do with the query being different since you are
>>> using different arguments. The result may be the same but dealing with the
>>> internals is very different, if you are familiar with relational database
>>> you know that the way you construct your query has a lot to do with the
>>> performance.
>>>
>>>
>>> That's kind of a vague response from Perforce.
>>> Did they say which one is better?
>>>
>>>
>>> Actually, I made a patch to change the p4poller.py behaviour. I used "p4
>>> changes -m1" to have the last revision and use the result output  to get
>>> the changelists range "p4 changes @%d, @%d" % (lastChange, headRevision)
>>>
>>>
>>> Does this improve the hit on your server?
>>>
>>> Thanks,
>>> Bill
>>>
>>>
>>> Regards
>>> Yasser
>>>
>>>
>>>
>>> 2013/9/27 William Deegan <bill at baddogconsulting.com>
>>>
>>>> Yasser,
>>>> On Sep 19, 2013, at 2:14 PM, Yasser LEBON <yasser.lebon at gmail.com>
>>>> wrote:
>>>>
>>>> >
>>>> > Hello,
>>>> >
>>>> > I noticed that buildbot use the following command to get perforce
>>>> depot list of changes since last sync:
>>>> >
>>>> > 'p4 changes %s...@%d,now' % (self.p4base, self.last_change+1)
>>>> >
>>>> > Is there any reason why we use 'now' instead of '#head'? Thanks
>>>>
>>>> I'm not sure why the previous maintainer used ,now, but the result
>>>> should be the same.
>>>>
>>>> From p4 docs:
>>>> * The datespec @now can be used as a synonym for the current date
>>>> and time.
>>>> * file#head The head revision (latest version) of file. Except where
>>>> explicitly
>>>> noted, this is equivalent to referring to the file without a revision
>>>> specifier.
>>>>
>>>> Are you aware of any functional difference?
>>>> -Bill
>>>
>>>
>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://buildbot.net/pipermail/devel/attachments/20131028/7c73b37b/attachment.html>