From noreply at pgfoundry.org  Sat Mar  8 18:56:00 2008
From: noreply at pgfoundry.org (noreply at pgfoundry.org)
Date: Sat,  8 Mar 2008 18:56:00 +0000 (UTC)
Subject: [ python-Bugs-1000295 ] Use Python to run documentation builds
Message-ID: <20080308185600.E701717ADD76@pgfoundry.org>

Bugs item #1000295, was opened at 2005-04-28 00:02
You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1000295&group_id=1000094

Category: Documentation
Group: To do
>Status: Closed
Resolution: Accepted
Priority: 3
Submitted By: James Pye (jwp)
Assigned to: Nobody (None)
Summary: Use Python to run documentation builds

Initial Comment:
Currently, the documentation is built with a Makefile, while this is fine, it would be nicer to have a Python based build script. The script could potentially  use Python bindings to do the build, maybe providing extra granularity and increasing portability.

----------------------------------------------------------------------

>Comment By: James Pye (jwp)
Date: 2008-03-08 11:56

Message:
Never mind this. The `doc` project is being deprecated in favor of rst files maintained within each project.

Quite relieving actually, as maintaining the documentation as I was was quite taxing. Also, a project driven structure will help to make sense of how things should layout.

----------------------------------------------------------------------

You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1000295&group_id=1000094

From noreply at pgfoundry.org  Thu Mar 13 23:43:48 2008
From: noreply at pgfoundry.org (noreply at pgfoundry.org)
Date: Thu, 13 Mar 2008 23:43:48 +0000 (UTC)
Subject: [ python-Bugs-1010316 ] interval format inconsistency
Message-ID: <20080313234348.21ECE17ADD86@pgfoundry.org>

Bugs item #1010316, was opened at 2008-03-13 16:43
You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1010316&group_id=1000094

Category: fe/proboscis
Group: None
Status: Open
Resolution: None
Priority: 5
Submitted By: James Pye (jwp)
Assigned to: James Pye (jwp)
Summary: interval format inconsistency

Initial Comment:
>>> ~query("SELECT timeofday()::timestamp - now()::timestamp")
Traceback (most recent call last):
  File "<console>", line 1, in <module>
  File "build/bdist.macosx-10.5-i386/egg/postgresql/interface/proboscis/tracenull.py", line 1147, in first
    return io(xt[0])
  File "build/bdist.macosx-10.5-i386/egg/postgresql/protocol/typical/stdio.py", line 92, in <lambda>
    lambda x: interval_unpack(ps.interval_unpack(x))
  File "build/bdist.macosx-10.5-i386/egg/postgresql/protocol/typical/pstruct.py", line 125, in interval_unpack
    tim, day, month = dll_unpack(data)
error: unpack requires a string argument of length 16

This works in 8.2, but not 8.0. I imagine the wire format changed. This will need to be a specific override in proboscis.

Both in 0.9.4 and 1.0dev.

----------------------------------------------------------------------

You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1010316&group_id=1000094

From noreply at pgfoundry.org  Fri Mar 14 14:37:15 2008
From: noreply at pgfoundry.org (noreply at pgfoundry.org)
Date: Fri, 14 Mar 2008 14:37:15 +0000 (UTC)
Subject: [ python-Bugs-1010316 ] interval format inconsistency
Message-ID: <20080314143715.05A0317AE698@pgfoundry.org>

Bugs item #1010316, was opened at 2008-03-13 16:43
You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1010316&group_id=1000094

Category: fe/proboscis
Group: None
Status: Open
Resolution: None
Priority: 5
Submitted By: James Pye (jwp)
Assigned to: James Pye (jwp)
Summary: interval format inconsistency

Initial Comment:
>>> ~query("SELECT timeofday()::timestamp - now()::timestamp")
Traceback (most recent call last):
  File "<console>", line 1, in <module>
  File "build/bdist.macosx-10.5-i386/egg/postgresql/interface/proboscis/tracenull.py", line 1147, in first
    return io(xt[0])
  File "build/bdist.macosx-10.5-i386/egg/postgresql/protocol/typical/stdio.py", line 92, in <lambda>
    lambda x: interval_unpack(ps.interval_unpack(x))
  File "build/bdist.macosx-10.5-i386/egg/postgresql/protocol/typical/pstruct.py", line 125, in interval_unpack
    tim, day, month = dll_unpack(data)
error: unpack requires a string argument of length 16

This works in 8.2, but not 8.0. I imagine the wire format changed. This will need to be a specific override in proboscis.

Both in 0.9.4 and 1.0dev.

----------------------------------------------------------------------

>Comment By: James Pye (jwp)
Date: 2008-03-14 07:37

Message:
I have a fix for this locally.

Simply, create another version of the interval packers and use dl_ instead of dll_. These new versions retain compatibility with the `days present` variants by zero-ing the days(ie, still takes and returns a triple, but the day is always zero).

Sadly, I'm either subjected to detecting the format by issuing an additional query on connect, or by version analysis(version_info <= 8,0). Currently, I'm leaning toward the version as I'd prefer to not add additional roundtrips to the cost of the startup.

----------------------------------------------------------------------

You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1010316&group_id=1000094

From noreply at pgfoundry.org  Fri Mar 14 17:35:22 2008
From: noreply at pgfoundry.org (noreply at pgfoundry.org)
Date: Fri, 14 Mar 2008 17:35:22 +0000 (UTC)
Subject: [ python-Bugs-1010221 ] Run interrupt on controller loss tracked by a
	weakref
Message-ID: <20080314173522.90CFF17AD0B1@pgfoundry.org>

Bugs item #1010221, was opened at 2008-01-05 22:35
You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1010221&group_id=1000094

Category: fe/proboscis
Group: None
Status: Open
Resolution: None
Priority: 4
Submitted By: James Pye (jwp)
Assigned to: James Pye (jwp)
Summary: Run interrupt on controller loss tracked by a weakref

Initial Comment:
With the new protocol transactions, it is possible to assign "controller" objects to a given transaction as a weakref. The weakref is necessary because it is important to know when the interface to the result set no longer exists. When an attempt to process further transactions is made, a check can be done on the weakref to identify if the controller still exists; if it doesn't an exception should be raised and the current transaction interrupted(via .Interrupt()) notifying the user of the potential data/state loss and for large copies interrupting a very costly completion. Normally these situations would occur  when an interrupt occurs(signal or keyboardinterrupt).

For the most part, this would normally see action on ResultHandles running COPYs, and Portals on the initial execute.

----------------------------------------------------------------------

>Comment By: James Pye (jwp)
Date: 2008-03-14 10:35

Message:
After some thought, this actually conflicts with the ability to push out multiple protocol transactions in order to reduce roundtrips.

The problem is that if you push multiple transactions, and an exception or other caused a lost controller, there is no way for the protocol transaction manager to immediately know what will be interrupted. However, if protocol transactions are done one at a time, it knows what will be interrupted, if anything.

So, as a result, I'm going to close out the multiple pushes task. While it's not dead, it's also not a 1.0 feature. More thought needs to be given to how it would work with lost controller garbage collection.

----------------------------------------------------------------------

You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1010221&group_id=1000094

From noreply at pgfoundry.org  Fri Mar 14 17:36:19 2008
From: noreply at pgfoundry.org (noreply at pgfoundry.org)
Date: Fri, 14 Mar 2008 17:36:19 +0000 (UTC)
Subject: [ python-Bugs-1010264 ] Implement a queue for protocol transactions
Message-ID: <20080314173619.462D117AD0B1@pgfoundry.org>

Bugs item #1010264, was opened at 2008-02-03 14:12
You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1010264&group_id=1000094

Category: fe/proboscis
Group: To do
>Status: Closed
>Resolution: Postponed
Priority: 5
Submitted By: James Pye (jwp)
Assigned to: James Pye (jwp)
Summary: Implement a queue for protocol transactions

Initial Comment:
Currently, only one transaction can be set for processing. However, some situations involving transactions can benefit substantially from queueing the transactions to avoid extra round-trips. Granted, any kind of commit does not have this luxury, but starts and rollbacks can certainly benefit from this.

This would allow request messages of a sequences of protocol transactions to be sent out in one write. Removing any need or desire to combine multiple requests into a single transaction.

----------------------------------------------------------------------

>Comment By: James Pye (jwp)
Date: 2008-03-14 10:36

Message:
Conflicts with lost controller garbage collection. This will is being deferred/closed for now.

----------------------------------------------------------------------

You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1010264&group_id=1000094

From noreply at pgfoundry.org  Mon Mar 17 14:51:48 2008
From: noreply at pgfoundry.org (noreply at pgfoundry.org)
Date: Mon, 17 Mar 2008 14:51:48 +0000 (UTC)
Subject: [ python-Bugs-1010316 ] interval format inconsistency
Message-ID: <20080317145148.156F317AE667@pgfoundry.org>

Bugs item #1010316, was opened at 2008-03-13 16:43
You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1010316&group_id=1000094

Category: fe/proboscis
Group: None
>Status: Closed
>Resolution: Accepted
Priority: 5
Submitted By: James Pye (jwp)
Assigned to: James Pye (jwp)
Summary: interval format inconsistency

Initial Comment:
>>> ~query("SELECT timeofday()::timestamp - now()::timestamp")
Traceback (most recent call last):
  File "<console>", line 1, in <module>
  File "build/bdist.macosx-10.5-i386/egg/postgresql/interface/proboscis/tracenull.py", line 1147, in first
    return io(xt[0])
  File "build/bdist.macosx-10.5-i386/egg/postgresql/protocol/typical/stdio.py", line 92, in <lambda>
    lambda x: interval_unpack(ps.interval_unpack(x))
  File "build/bdist.macosx-10.5-i386/egg/postgresql/protocol/typical/pstruct.py", line 125, in interval_unpack
    tim, day, month = dll_unpack(data)
error: unpack requires a string argument of length 16

This works in 8.2, but not 8.0. I imagine the wire format changed. This will need to be a specific override in proboscis.

Both in 0.9.4 and 1.0dev.

----------------------------------------------------------------------

>Comment By: James Pye (jwp)
Date: 2008-03-17 07:51

Message:
Checked into HEAD. Merge to 0.9 pending.

----------------------------------------------------------------------

Comment By: James Pye (jwp)
Date: 2008-03-14 07:37

Message:
I have a fix for this locally.

Simply, create another version of the interval packers and use dl_ instead of dll_. These new versions retain compatibility with the `days present` variants by zero-ing the days(ie, still takes and returns a triple, but the day is always zero).

Sadly, I'm either subjected to detecting the format by issuing an additional query on connect, or by version analysis(version_info <= 8,0). Currently, I'm leaning toward the version as I'd prefer to not add additional roundtrips to the cost of the startup.

----------------------------------------------------------------------

You can respond by visiting: 
http://pgfoundry.org/tracker/?func=detail&atid=442&aid=1010316&group_id=1000094