You are viewing a plain text version of this content. The canonical link for it is here.
Posted to derby-dev@db.apache.org by Myrna van Lunteren <m....@gmail.com> on 2011/09/13 06:56:52 UTC

10.8.2.0 release candidate generated...but...

Hi,

I had to do some tweaking of/around the ant release build target
(about which I will worry tomorrow), but I did manage to get the RC1
for 10.8.2.0 on the public_html site.

However, the automated process set the drdamaint number back to 0, and
I suspect that was wrong. So, before calling the test and vote I'd
like to make sure the fix for DERBY-5236 is still working correctly.
Otherwise I will need to spin a RC2 after looking at the release
target to prevent it from resetting that number.

I'm running tests, but I didn't see any specific test included in the
backport, so, Knut, can you check the release works properly with
regard to your fix?

It can be downloaded from:
http://people.apache.org/~myrnavl/derby-10.8.2.0-RC1/

Thx,
Myrna

Re: 10.8.2.0 release candidate generated...but...

Posted by Myrna van Lunteren <m....@gmail.com>.
On Tue, Sep 13, 2011 at 4:20 AM, Knut Anders Hatlen
<kn...@oracle.com> wrote:
> Myrna van Lunteren <m....@gmail.com> writes:
>
>> I'm running tests, but I didn't see any specific test included in the
>> backport, so, Knut, can you check the release works properly with
>> regard to your fix?
>
> Hi Myrna,
>
> There is a test for DERBY-5236 in derbynet.PrepareStatementTest, and it
> fails with RC1:
>
> There was 1 failure:
> 1) testLongColumn(org.apache.derbyTesting.functionTests.tests.derbynet.PrepareStatementTest)junit.framework.AssertionFailedError: Column value mismatch @ col
> umn '1', row 1:
>    Expected: >丐丐丐丐丐丐丐丐...
>    Found:    >丐丐丐丐丐丐丐丐...
>        at org.apache.derbyTesting.junit.JDBC.assertRowInResultSet(JDBC.java:1213)
>        at org.apache.derbyTesting.junit.JDBC.assertRowInResultSet(JDBC.java:1125)
> (...)
>
> Setting the DRDA maintenance version back to 1 should make it pass
> again.
>
> --
> Knut Anders
>

Thanks for checking on this. I'll spin a new version.

And re download permissions - that's right, I had to use sftp (not
sure why scp wasn't working for me) and I checked on download after
the first couple - then forgot to do the others. Anyway, I'll soon
blow that directory away, no point keeping it. I'll make sure to chmod
the RC2.

Myrna

Re: 10.8.2.0 release candidate generated...but...

Posted by Knut Anders Hatlen <kn...@oracle.com>.
Myrna van Lunteren <m....@gmail.com> writes:

> I'm running tests, but I didn't see any specific test included in the
> backport, so, Knut, can you check the release works properly with
> regard to your fix?

Hi Myrna,

There is a test for DERBY-5236 in derbynet.PrepareStatementTest, and it
fails with RC1:

There was 1 failure:
1) testLongColumn(org.apache.derbyTesting.functionTests.tests.derbynet.PrepareStatementTest)junit.framework.AssertionFailedError: Column value mismatch @ col
umn '1', row 1:
    Expected: >丐丐丐丐丐丐丐丐...
    Found:    >丐丐丐丐丐丐丐丐...
        at org.apache.derbyTesting.junit.JDBC.assertRowInResultSet(JDBC.java:1213)
        at org.apache.derbyTesting.junit.JDBC.assertRowInResultSet(JDBC.java:1125)
(...)

Setting the DRDA maintenance version back to 1 should make it pass
again.

-- 
Knut Anders

Re: 10.8.2.0 release candidate generated...but...

Posted by Kathey Marsden <km...@sbcglobal.net>.
On 9/13/2011 6:37 AM, Rick Hillegas wrote:
>>> org.apache.derbyPreBuild.ReleaseProperties. Right now that program 
>>> hardcodes
>>> drdamaint to be 0. Here are some ways we can change that program:
>>>
>>> 1) Hardcode drdamaint to be 1.
>>>
>>> 2) Make drdamaint a parameter to the program, like the Derby version 
>>> id.
>>>
>>> 3) Make the program retrieve drdamaint from the existing 
>>> release.properties
>>> file and write exactly what was read.
>>>
Since releases will always be made off a branch, I think option 3 makes 
the most sense.  It is always fine I think for it to be 0 on the trunk 
as we don't normally track protocol  changes within the development 
line  of a major release.

  The release level communicated in the PRDID and used to communicate 
the DDM server level and make protocol decisions has two digits for the 
version level, two digits for the release level  and a single digit  for 
a modify level which corresponds to our drda maintence version.  It can  
be bumped when ever protocol fixes need to be backported to a branch 
and  can be reset with each new minor (second digit) release

As I recall, the original reason modifyLevel  was not made to correspond 
to the third digit  of our version was that way back when Network Server 
was created there were only three parts to the version and the third 
digit was bumped frequently, so for example we would have versions like 
5.1.45. Since there was not enough real estate for the actual third part 
of the version  we just added this rarely used DRDA maintenance version 
to allow protocol fixes to occur in the branches.

I hope that helps and also hope that I remember it correctly.

Kathey



Re: 10.8.2.0 release candidate generated...but...

Posted by Rick Hillegas <ri...@oracle.com>.
On 9/13/11 6:25 AM, Myrna van Lunteren wrote:
> On Tue, Sep 13, 2011 at 6:10 AM, Rick Hillegas<ri...@oracle.com>  wrote:
>> On 9/12/11 9:56 PM, Myrna van Lunteren wrote:
>>> Hi,
>>>
>>> I had to do some tweaking of/around the ant release build target
>>> (about which I will worry tomorrow), but I did manage to get the RC1
>>> for 10.8.2.0 on the public_html site.
>>>
>>> However, the automated process set the drdamaint number back to 0, and
>>> I suspect that was wrong.
>> The release properties are written by
>> org.apache.derbyPreBuild.ReleaseProperties. Right now that program hardcodes
>> drdamaint to be 0. Here are some ways we can change that program:
>>
>> 1) Hardcode drdamaint to be 1.
>>
>> 2) Make drdamaint a parameter to the program, like the Derby version id.
>>
>> 3) Make the program retrieve drdamaint from the existing release.properties
>> file and write exactly what was read.
>>
>> The meaning of drdamaint is unclear to me and I don't understand why
>> drdamaint is 0 on the trunk but is supposed to be 1 on the 10.8 branch.
>>
>> What are your thoughts about the meaning of drdamaint and how this program
>> should behave?
>>
>> Thanks,
>> -Rick
>>
>>
>>>   So, before calling the test and vote I'd
>>> like to make sure the fix for DERBY-5236 is still working correctly.
>>> Otherwise I will need to spin a RC2 after looking at the release
>>> target to prevent it from resetting that number.
>>>
>>> I'm running tests, but I didn't see any specific test included in the
>>> backport, so, Knut, can you check the release works properly with
>>> regard to your fix?
>>>
>>> It can be downloaded from:
>>> http://people.apache.org/~myrnavl/derby-10.8.2.0-RC1/
>>>
>>> Thx,
>>> Myrna
>>>
>>
> I don't understand the details either or I would've fixed it right
> away - I thought perhaps it worked in combination with the 3rd digit -
> but apparently not. So perhaps the drdamaint version needs to be upped
> on trunk also?
>
> I would like it best if the release process could be automated enough
> to get the old drdamaint version from the release.properties file. I
> could do that now, we can just put it as an action under DERBY-5287,
> but if you have some ideas on how to achieve it, that would help.
I don't have any clever ideas about how to do this. There is already a 
handle on the release properties file used by the execute() method. The 
file handle is called target. Should be possible to just open a 
java.io.LineNumberReader on that file, loop through the lines looking 
for one that starts with "drdamaint=", save that line in a variable, and 
just re-emit it in the execute() method where the hardcoded drdamaint is 
currently set.

Thanks,
-Rick
> Myrna
>


Re: 10.8.2.0 release candidate generated...but...

Posted by Knut Anders Hatlen <kn...@oracle.com>.
Myrna van Lunteren <m....@gmail.com> writes:

> I don't understand the details either or I would've fixed it right
> away - I thought perhaps it worked in combination with the 3rd digit -
> but apparently not.

It works in combination with the first two digits in the version number,
but it is less significant than those two. So the tuple (10, 9, 0) on
trunk is higher than (10, 8, 1) on the 10.8 branch, even if 0 is lower
than 1.

> So perhaps the drdamaint version needs to be upped
> on trunk also?

My understanding is that it should be set on the branch only.

> I would like it best if the release process could be automated enough
> to get the old drdamaint version from the release.properties file.

Yes, I think keeping the old drdamaint version is the correct behaviour
of the release script. We typically don't want to change that property
when we bump the version, only when we backport a change that needs
special handling in the protocol.

Changing it manually in release.properties should be enough to get it
working again for now, though.

-- 
Knut Anders

Re: 10.8.2.0 release candidate generated...but...

Posted by Myrna van Lunteren <m....@gmail.com>.
On Tue, Sep 13, 2011 at 6:10 AM, Rick Hillegas <ri...@oracle.com> wrote:
> On 9/12/11 9:56 PM, Myrna van Lunteren wrote:
>>
>> Hi,
>>
>> I had to do some tweaking of/around the ant release build target
>> (about which I will worry tomorrow), but I did manage to get the RC1
>> for 10.8.2.0 on the public_html site.
>>
>> However, the automated process set the drdamaint number back to 0, and
>> I suspect that was wrong.
>
> The release properties are written by
> org.apache.derbyPreBuild.ReleaseProperties. Right now that program hardcodes
> drdamaint to be 0. Here are some ways we can change that program:
>
> 1) Hardcode drdamaint to be 1.
>
> 2) Make drdamaint a parameter to the program, like the Derby version id.
>
> 3) Make the program retrieve drdamaint from the existing release.properties
> file and write exactly what was read.
>
> The meaning of drdamaint is unclear to me and I don't understand why
> drdamaint is 0 on the trunk but is supposed to be 1 on the 10.8 branch.
>
> What are your thoughts about the meaning of drdamaint and how this program
> should behave?
>
> Thanks,
> -Rick
>
>
>>  So, before calling the test and vote I'd
>> like to make sure the fix for DERBY-5236 is still working correctly.
>> Otherwise I will need to spin a RC2 after looking at the release
>> target to prevent it from resetting that number.
>>
>> I'm running tests, but I didn't see any specific test included in the
>> backport, so, Knut, can you check the release works properly with
>> regard to your fix?
>>
>> It can be downloaded from:
>> http://people.apache.org/~myrnavl/derby-10.8.2.0-RC1/
>>
>> Thx,
>> Myrna
>>
>
>

I don't understand the details either or I would've fixed it right
away - I thought perhaps it worked in combination with the 3rd digit -
but apparently not. So perhaps the drdamaint version needs to be upped
on trunk also?

I would like it best if the release process could be automated enough
to get the old drdamaint version from the release.properties file. I
could do that now, we can just put it as an action under DERBY-5287,
but if you have some ideas on how to achieve it, that would help.

Myrna

Re: 10.8.2.0 release candidate generated...but...

Posted by Rick Hillegas <ri...@oracle.com>.
On 9/12/11 9:56 PM, Myrna van Lunteren wrote:
> Hi,
>
> I had to do some tweaking of/around the ant release build target
> (about which I will worry tomorrow), but I did manage to get the RC1
> for 10.8.2.0 on the public_html site.
>
> However, the automated process set the drdamaint number back to 0, and
> I suspect that was wrong.
The release properties are written by 
org.apache.derbyPreBuild.ReleaseProperties. Right now that program 
hardcodes drdamaint to be 0. Here are some ways we can change that program:

1) Hardcode drdamaint to be 1.

2) Make drdamaint a parameter to the program, like the Derby version id.

3) Make the program retrieve drdamaint from the existing 
release.properties file and write exactly what was read.

The meaning of drdamaint is unclear to me and I don't understand why 
drdamaint is 0 on the trunk but is supposed to be 1 on the 10.8 branch.

What are your thoughts about the meaning of drdamaint and how this 
program should behave?

Thanks,
-Rick


>   So, before calling the test and vote I'd
> like to make sure the fix for DERBY-5236 is still working correctly.
> Otherwise I will need to spin a RC2 after looking at the release
> target to prevent it from resetting that number.
>
> I'm running tests, but I didn't see any specific test included in the
> backport, so, Knut, can you check the release works properly with
> regard to your fix?
>
> It can be downloaded from:
> http://people.apache.org/~myrnavl/derby-10.8.2.0-RC1/
>
> Thx,
> Myrna
>


Re: 10.8.2.0 release candidate generated...but...

Posted by Kristian Waagan <kr...@oracle.com>.
On 13.09.11 06:56, Myrna van Lunteren wrote:
> Hi,
>
> I had to do some tweaking of/around the ant release build target
> (about which I will worry tomorrow), but I did manage to get the RC1
> for 10.8.2.0 on the public_html site.
>
> However, the automated process set the drdamaint number back to 0, and
> I suspect that was wrong. So, before calling the test and vote I'd
> like to make sure the fix for DERBY-5236 is still working correctly.
> Otherwise I will need to spin a RC2 after looking at the release
> target to prevent it from resetting that number.
>
> I'm running tests, but I didn't see any specific test included in the
> backport, so, Knut, can you check the release works properly with
> regard to your fix?
>
> It can be downloaded from:
> http://people.apache.org/~myrnavl/derby-10.8.2.0-RC1/

Hi Myrna,

Some of the files in that directory are lacking the required read 
privilege and cannot be downloaded. This one does work though:
http://people.apache.org/~myrnavl/derby-10.8.2.0-RC1/db-derby-10.8.2.0-bin.zip

Also, I backported a test fix (DERBY-5394) to the 10.8 branch.
Currently, there's a high chance that running 'ant junit-lowmem' will 
fail with RC1.


-- 
Kristian

>
> Thx,
> Myrna