Re: Maintenance form exection thread

Le 10/04/2010 11:14, Magnus Hagander a écrit :
> On Sat, Apr 10, 2010 at 10:58, Guillaume Lelarge <guillaume(at)lelarge(dot)info> wrote:
>> Le 06/04/2010 22:33, Guillaume Lelarge a écrit :
>>> Le 06/04/2010 21:48, Magnus Hagander a écrit :
>>>> On Tue, Apr 6, 2010 at 21:01, Dave Page <dpage(at)pgadmin(dot)org> wrote:
>>>>> On Tue, Apr 6, 2010 at 6:11 PM, Magnus Hagander <magnus(at)hagander(dot)net> wrote:
>>>>>> Hi!
>>>>>>
>>>>>> Am I reading things right that we actually execute things from the
>>>>>> maintenance dialog (like VACUUM) on a separate thread, to keep the UI
>>>>>> responsive? Yet, it keeps hanging when we do that. My guess is that
>>>>>> we're "using up" the connection we have, and as soon as somebody else
>>>>>> needs access to the connection to do things like refresh tree
>>>>>> information, we hang and wait. Or am I reading this wrong?
>>>>>>
>>>>>> If we are, should we perhaps consider firing off these jobs on a
>>>>>> separate connection?
>>>>>
>>>>> Yeah, that would seem like a sensible idea. At first thought I guessed
>>>>> it was an issue like this
>>>>> http://svn.pgadmin.org/cgi-bin/viewcvs.cgi?rev=6458&view=rev, but on
>>>>> reflection I think your much more simple explanation is the likely
>>>>> cause.
>>>>
>>>> I don' t have time to look into it right onw. Do you, or should I just
>>>> add a ticket for it for "eventual fixing"?
>>>>
>>>
>>> Add the ticket, I will take care of it this week if no one does.
>>>
>>
>> Here is a patch for trunk.
>>
>> Oh, and one question. You created a *bug* ticket. Do you mean you want
>> this applied on the 1.10 branch? I prefer to ask as I don't really think
>> this is a bug, it's more of an enhancement to me.
>
> I do consider it a bug.
>
> If it's backpatchable or not depends on the patch complexity, imo.

Applies good, needs to fix the CreateConn call (because we don't support
the application_name in 1.10), compiles great then, and works great too.

> Given that the solution is creating a separate connection for it, I
> think it should *not* be applied to 1.10, because it's a large
> problem. If someone had corrected my diagnosis and found a
> lower-impact way, then it could've been.
>
> The patch looks surprisingly simple :-) But I can't see why it
> wouldn't be correct - looks good to me.
>

OK, will apply to trunk.

--
Guillaume.
http://www.postgresqlfr.org
http://dalibo.com