10.8.2.2 coming up

[Myrna van Lunteren <m....@gmail.com>]

Hi,

Rick has backed out the fix for DERBY-4437.

I'm running nstest on a couple of machines and - of course depending
on the results in the next day or so - I hope to make a new candidate
release on Monday, October 10, 2011.

I'm not updating the wiki pages just yet...

I believe most fixed bugs that are suitable for backport have been backported.

I only have the following possibles:
DERBY-3823 - NullPointerException in stress.multi test
      The fix hasn't yet been checked in, but this might be nice to
avoid this old intermittent test failure. Is this change suitable for
backport or could it cause incompatibilities?

DERBY-5427 - I think Kathey is on this. Would be nice to have, but it
only affects ibm 1.7 users and only in specific situations.

Any issues anyone is working on backporting?

Regards,
Myrna

Re: 10.8.2.2 coming up

[Rick Hillegas <ri...@oracle.com>]

On 10/10/11 12:07 PM, Myrna van Lunteren wrote:
> On Mon, Oct 10, 2011 at 1:01 AM, Kristian Waagan
> <kr...@oracle.com>  wrote:
>> On 08.10.11 03:31, Myrna van Lunteren wrote:
>>> On Fri, Oct 7, 2011 at 2:10 AM, Kristian Waagan
>>> <kr...@oracle.com>    wrote:
>>>> Den 07.10.2011 00:36, skrev Myrna van Lunteren:
>> <  snip>
>>
>>>>> Any issues anyone is working on backporting?
>>>> Hi  Myrna,
>>>>
>>>> I'm working on DERBY-5447, which is a Java deadlock involving the
>>>> automatic
>>>> index statistics daemon. It causes two threads to hang indefinitely.
>>>> I have two patches, one ran for 1099 iterations of the hanging test suite
>>>> without hitting the problem, the other is currently at 879 iterations
>>>> (each
>>>> iteration takes 60 - 90 seconds). Without the fix(es) the test ususally
>>>> hangs after 9 to 30 iterations.
>>>>
>>>> The good news is that the second patch is an isolated change to the istat
>>>> daemon itself, instead of modifying the core code in
>>>> BasePage.releaseExclusive. I wouldn't be afraid to port this to 10.8
>>>> immediately after the fix goes into trunk.
>>>>
>>>> I don't want you to hold back the release because of this issue (as
>>>> several
>>>> factors have to come into play to trigger it), but it would be good to
>>>> know
>>>> when you plan to produce the next release candidate.
>>>>
>>>>
>>>> Regards, and thank you for your continued efforts for a solid 10.8.2,
>>>> --
>>>> Kristian
>>>>
>>>>> Regards,
>>>>> Myrna
>>>>
>>> I really wanted to make a release candidate on Monday, but I think
>>> it's not useful while we're analyzing various error conditions after
>>> Rick's backing out work for DERBY-5448.
>>> So, at this point, I don't think I'll make a release candidate on
>>> Monday. Let's see where we are at the end of Monday.
>> Ok.
>> In any case, I backported DERBY-5447 to 10.8 with revision 1180811.
>>
>>
>> --
>> Kristian
>>
>>> Myrna
>>
> Thank you all for your feedback.
>
> It seems to me that apart from another deadlock situation which Rick
I regard the evidence as inconclusive. So far no one has reproduced the 
identity generator deadlock in 10.8.1. There could still be a regression 
introduced by 10.8.2. We don't understand the problem yet.
> and Mike suspect is pre-existing 10.8.2 and a worrying but likely
> unreproducible error (DERBY-5450) there is nothing more to be done for
> this release.
>
> I'll spin a release candidate now.
>
> Myrna
>

Re: 10.8.2.2 coming up

[Rick Hillegas <ri...@oracle.com>]

On 10/10/11 12:27 PM, Myrna van Lunteren wrote:
>> Thank you all for your feedback.
>>
>> It seems to me that apart from another deadlock situation which Rick
>> and Mike suspect is pre-existing 10.8.2 and a worrying but likely
>> unreproducible error (DERBY-5450) there is nothing more to be done for
>> this release.
>>
>> I'll spin a release candidate now.
>>
>> Myrna
>>
> But, what to do with the release notes regarding for DERBY-4437?  If I
> do my filter with 'resolved in' versions 10.8.2.0 and 10.8.2.1 and
> 10.8.2.2, it will show up as fixed, however, it's been backed out, so
> not fixed.
>
> Suggestions?
Hi Myrna,

I have removed 10.8.2 from the "fixed in" list for DERBY-4437. That may 
remove the JIRA from the release notes.

Thanks,
-Rick
> Myrna
>

Re: 10.8.2.2 coming up

[Myrna van Lunteren <m....@gmail.com>]

> Thank you all for your feedback.
>
> It seems to me that apart from another deadlock situation which Rick
> and Mike suspect is pre-existing 10.8.2 and a worrying but likely
> unreproducible error (DERBY-5450) there is nothing more to be done for
> this release.
>
> I'll spin a release candidate now.
>
> Myrna
>

But, what to do with the release notes regarding for DERBY-4437?  If I
do my filter with 'resolved in' versions 10.8.2.0 and 10.8.2.1 and
10.8.2.2, it will show up as fixed, however, it's been backed out, so
not fixed.

Suggestions?

Myrna

Re: 10.8.2.2 coming up

[Myrna van Lunteren <m....@gmail.com>]

On Mon, Oct 10, 2011 at 1:01 AM, Kristian Waagan
<kr...@oracle.com> wrote:
> On 08.10.11 03:31, Myrna van Lunteren wrote:
>>
>> On Fri, Oct 7, 2011 at 2:10 AM, Kristian Waagan
>> <kr...@oracle.com>  wrote:
>>>
>>> Den 07.10.2011 00:36, skrev Myrna van Lunteren:
>
> < snip >
>
>>>> Any issues anyone is working on backporting?
>>>
>>> Hi  Myrna,
>>>
>>> I'm working on DERBY-5447, which is a Java deadlock involving the
>>> automatic
>>> index statistics daemon. It causes two threads to hang indefinitely.
>>> I have two patches, one ran for 1099 iterations of the hanging test suite
>>> without hitting the problem, the other is currently at 879 iterations
>>> (each
>>> iteration takes 60 - 90 seconds). Without the fix(es) the test ususally
>>> hangs after 9 to 30 iterations.
>>>
>>> The good news is that the second patch is an isolated change to the istat
>>> daemon itself, instead of modifying the core code in
>>> BasePage.releaseExclusive. I wouldn't be afraid to port this to 10.8
>>> immediately after the fix goes into trunk.
>>>
>>> I don't want you to hold back the release because of this issue (as
>>> several
>>> factors have to come into play to trigger it), but it would be good to
>>> know
>>> when you plan to produce the next release candidate.
>>>
>>>
>>> Regards, and thank you for your continued efforts for a solid 10.8.2,
>>> --
>>> Kristian
>>>
>>>>
>>>> Regards,
>>>> Myrna
>>>
>>>
>>
>> I really wanted to make a release candidate on Monday, but I think
>> it's not useful while we're analyzing various error conditions after
>> Rick's backing out work for DERBY-5448.
>> So, at this point, I don't think I'll make a release candidate on
>> Monday. Let's see where we are at the end of Monday.
>
> Ok.
> In any case, I backported DERBY-5447 to 10.8 with revision 1180811.
>
>
> --
> Kristian
>
>>
>> Myrna
>
>
Thank you all for your feedback.

It seems to me that apart from another deadlock situation which Rick
and Mike suspect is pre-existing 10.8.2 and a worrying but likely
unreproducible error (DERBY-5450) there is nothing more to be done for
this release.

I'll spin a release candidate now.

Myrna

Re: 10.8.2.2 coming up

[Kristian Waagan <kr...@oracle.com>]

On 08.10.11 03:31, Myrna van Lunteren wrote:
> On Fri, Oct 7, 2011 at 2:10 AM, Kristian Waagan
> <kr...@oracle.com>  wrote:
>> Den 07.10.2011 00:36, skrev Myrna van Lunteren:

< snip >

>>> Any issues anyone is working on backporting?
>>
>> Hi  Myrna,
>>
>> I'm working on DERBY-5447, which is a Java deadlock involving the automatic
>> index statistics daemon. It causes two threads to hang indefinitely.
>> I have two patches, one ran for 1099 iterations of the hanging test suite
>> without hitting the problem, the other is currently at 879 iterations (each
>> iteration takes 60 - 90 seconds). Without the fix(es) the test ususally
>> hangs after 9 to 30 iterations.
>>
>> The good news is that the second patch is an isolated change to the istat
>> daemon itself, instead of modifying the core code in
>> BasePage.releaseExclusive. I wouldn't be afraid to port this to 10.8
>> immediately after the fix goes into trunk.
>>
>> I don't want you to hold back the release because of this issue (as several
>> factors have to come into play to trigger it), but it would be good to know
>> when you plan to produce the next release candidate.
>>
>>
>> Regards, and thank you for your continued efforts for a solid 10.8.2,
>> --
>> Kristian
>>
>>>
>>> Regards,
>>> Myrna
>>
>>
>
> I really wanted to make a release candidate on Monday, but I think
> it's not useful while we're analyzing various error conditions after
> Rick's backing out work for DERBY-5448.
> So, at this point, I don't think I'll make a release candidate on
> Monday. Let's see where we are at the end of Monday.

Ok.
In any case, I backported DERBY-5447 to 10.8 with revision 1180811.


-- 
Kristian

>
> Myrna

Re: 10.8.2.2 coming up

[Myrna van Lunteren <m....@gmail.com>]

On Fri, Oct 7, 2011 at 2:10 AM, Kristian Waagan
<kr...@oracle.com> wrote:
> Den 07.10.2011 00:36, skrev Myrna van Lunteren:
>>
>> Hi,
>>
>> Rick has backed out the fix for DERBY-4437.
>>
>> I'm running nstest on a couple of machines and - of course depending
>> on the results in the next day or so - I hope to make a new candidate
>> release on Monday, October 10, 2011.
>>
>> I'm not updating the wiki pages just yet...
>>
>> I believe most fixed bugs that are suitable for backport have been
>> backported.
>>
>> I only have the following possibles:
>> DERBY-3823 - NullPointerException in stress.multi test
>>       The fix hasn't yet been checked in, but this might be nice to
>> avoid this old intermittent test failure. Is this change suitable for
>> backport or could it cause incompatibilities?
>>
>> DERBY-5427 - I think Kathey is on this. Would be nice to have, but it
>> only affects ibm 1.7 users and only in specific situations.
>>
>> Any issues anyone is working on backporting?
>
> Hi  Myrna,
>
> I'm working on DERBY-5447, which is a Java deadlock involving the automatic
> index statistics daemon. It causes two threads to hang indefinitely.
> I have two patches, one ran for 1099 iterations of the hanging test suite
> without hitting the problem, the other is currently at 879 iterations (each
> iteration takes 60 - 90 seconds). Without the fix(es) the test ususally
> hangs after 9 to 30 iterations.
>
> The good news is that the second patch is an isolated change to the istat
> daemon itself, instead of modifying the core code in
> BasePage.releaseExclusive. I wouldn't be afraid to port this to 10.8
> immediately after the fix goes into trunk.
>
> I don't want you to hold back the release because of this issue (as several
> factors have to come into play to trigger it), but it would be good to know
> when you plan to produce the next release candidate.
>
>
> Regards, and thank you for your continued efforts for a solid 10.8.2,
> --
> Kristian
>
>>
>> Regards,
>> Myrna
>
>

I really wanted to make a release candidate on Monday, but I think
it's not useful while we're analyzing various error conditions after
Rick's backing out work for DERBY-5448.
So, at this point, I don't think I'll make a release candidate on
Monday. Let's see where we are at the end of Monday.

Myrna

Re: 10.8.2.2 coming up

[Kristian Waagan <kr...@oracle.com>]

Den 07.10.2011 00:36, skrev Myrna van Lunteren:
> Hi,
>
> Rick has backed out the fix for DERBY-4437.
>
> I'm running nstest on a couple of machines and - of course depending
> on the results in the next day or so - I hope to make a new candidate
> release on Monday, October 10, 2011.
>
> I'm not updating the wiki pages just yet...
>
> I believe most fixed bugs that are suitable for backport have been backported.
>
> I only have the following possibles:
> DERBY-3823 - NullPointerException in stress.multi test
>        The fix hasn't yet been checked in, but this might be nice to
> avoid this old intermittent test failure. Is this change suitable for
> backport or could it cause incompatibilities?
>
> DERBY-5427 - I think Kathey is on this. Would be nice to have, but it
> only affects ibm 1.7 users and only in specific situations.
>
> Any issues anyone is working on backporting?

Hi  Myrna,

I'm working on DERBY-5447, which is a Java deadlock involving the 
automatic index statistics daemon. It causes two threads to hang 
indefinitely.
I have two patches, one ran for 1099 iterations of the hanging test 
suite without hitting the problem, the other is currently at 879 
iterations (each iteration takes 60 - 90 seconds). Without the fix(es) 
the test ususally hangs after 9 to 30 iterations.

The good news is that the second patch is an isolated change to the 
istat daemon itself, instead of modifying the core code in 
BasePage.releaseExclusive. I wouldn't be afraid to port this to 10.8 
immediately after the fix goes into trunk.

I don't want you to hold back the release because of this issue (as 
several factors have to come into play to trigger it), but it would be 
good to know when you plan to produce the next release candidate.


Regards, and thank you for your continued efforts for a solid 10.8.2,
-- 
Kristian

>
> Regards,
> Myrna

Re: 10.8.2.2 coming up

["Dag H. Wanvik" <da...@oracle.com>]

Hi Myrna!

On 10/7/2011 12:36 AM, Myrna van Lunteren wrote:
> I only have the following possibles:
> DERBY-3823 - NullPointerException in stress.multi test
>        The fix hasn't yet been checked in, but this might be nice to
> avoid this old intermittent test failure. Is this change suitable for
> backport or could it cause incompatibilities?

I don't believe it should cause any incompatibilities (provided it's 
correct ;-)
Since it touches a central piece of code, I would not recommend 
backporting it till it has been reviewed and we have some confidence in it.
Any volunteers? ;-)

Thanks,
Dag
> DERBY-5427 - I think Kathey is on this. Would be nice to have, but it
> only affects ibm 1.7 users and only in specific situations.
>
> Any issues anyone is working on backporting?
>
> Regards,
> Myrna

Re: 10.8.2.2 coming up

[Mike Matrigali <mi...@sbcglobal.net>]

Great, I think it is time to have a new RC, given that a number
of blockers for the previous RC that was canceled have been checked in. 
  This way
everyone running tests can be using the same proposed release.

At this point I don't think the deadlock issues are worth holding
up a release for, though I would still hope they are high on
everyone's radar.

Does anyone consider anything in the current 10.8 branch a blocker
for a release at this time, assuming a RC made at this time passes
community testing?

/mikem

Myrna van Lunteren wrote:
> Hi,
> 
> Rick has backed out the fix for DERBY-4437.
> 
> I'm running nstest on a couple of machines and - of course depending
> on the results in the next day or so - I hope to make a new candidate
> release on Monday, October 10, 2011.
> 
> I'm not updating the wiki pages just yet...
> 
> I believe most fixed bugs that are suitable for backport have been backported.
> 
> I only have the following possibles:
> DERBY-3823 - NullPointerException in stress.multi test
>       The fix hasn't yet been checked in, but this might be nice to
> avoid this old intermittent test failure. Is this change suitable for
> backport or could it cause incompatibilities?
> 
> DERBY-5427 - I think Kathey is on this. Would be nice to have, but it
> only affects ibm 1.7 users and only in specific situations.
> 
> Any issues anyone is working on backporting?
> 
> Regards,
> Myrna
> 

Re: 10.8.2.2 coming up

[Kathey Marsden <km...@sbcglobal.net>]

On 10/6/2011 3:36 PM, Myrna van Lunteren wrote:
>
> DERBY-5427 - I think Kathey is on this. Would be nice to have, but it
> only affects ibm 1.7 users and only in specific situations.
I think this affects JDK 1.5 or higher for all vendors  to the extent 
that the thread dump will print to the derby.log.  For IBM  it is worse 
because of the javacores.  I just started looking at it and  don't have 
any insight yet as to how it might be fixed. I am going to work first on 
a smaller test case that I will post to the issue.  I thought just a 
shutdown authentication error would do it, but that didn't seem to be 
the case. I'll work hard on it this weekend but can't promise for 
Monday. Stay tuned ....

Thanks

Kathey