durusmail: qp: specifying host for scgi, durus, etc.
specifying host for scgi, durus, etc.
2006-03-30
2006-03-30
Re: specifying host for scgi, durus, etc.
2006-03-31
2006-03-31
2006-03-31
2006-03-31
2006-03-31
2006-03-31
2006-04-01
2006-04-01
2006-04-01
2006-04-01
2006-04-01
2006-04-01
2006-04-01
2006-04-02
specifying host for scgi, durus, etc.
mario ruggier
2006-04-02
On Apr 1, 2006, at 10:08 PM, David Binger wrote:
> On Apr 1, 2006, at 12:01 PM, mario ruggier wrote:
>> On Apr 1, 2006, at 11:07 AM, mario ruggier wrote:
>>
>>>     def process_hit(self, hit):
>>>         try:
>>>             self.get_connection().abort()
>>
>> Doing this on each hit loses the unsaved session data, so breaks
>> login pretty badly. Calling commit() instead seems to function ok,
>> but may not be the best thing to do? What would be the better thing
>> to call here to force a socket.error if server is down or connection
>> is broken?
>
> Every request response cycle should end with either
> a commit or abort, so there should *never* be any
> unsaved data around when process_hit is called.
> This is here to invalidate any
> oids that other clients have changed since since the
> last commit/abort called by *this* client.

Ah, good. I figured out what was happening in my case... some confusion
about which super process_hit() was being called, and I was
inadvertently bypassing the one on DurusPublisher. Thus, replacing the
last line of the process_hit() I posted earlier, i.e:
         qp.pub.publish.Publisher.process_hit(self, hit)
with:
         super(Publisher, self).process_hit(hit)

makes it all behave as expected.
So, I switch back to using abort()...

reply