Hey guys,

I'm trying to get the bug fixes for
https://issues.apache.org/jira/browse/ZOOKEEPER-1046 and
https://issues.apache.org/jira/browse/ZOOKEEPER-1069 ported back to
the 3.3 branch, but the tests written for these fixes won't port
because the test case was never checked in to that branch. What should
I do here? Check in the fixes to the branch with no test? I'm not
entirely crazy about that option but it seems like that has been done
before (the test in question was originally created for a different
bug fix that was ported without tests to 3.3 from trunk).
I'm ok to do that for these fixes but going forward do we really want
to be pushing fixes in anywhere without tests? I'd hate to have a
"fix" that doesn't work and we don't catch due to this.

Thanks,
C

yeah, branches are a pain to maintain. we have done both ways: checked
in the patch without the test and also backported the test. the later
is the preferable one in my opinion. the test case is a whole new
class right? can we just copy the class from trunk?

ben

On Sat, May 28, 2011 at 11:22 AM, Camille Fournier <[EMAIL PROTECTED]> wrote:
> Hey guys,
>
> I'm trying to get the bug fixes for
> https://issues.apache.org/jira/browse/ZOOKEEPER-1046 and
> https://issues.apache.org/jira/browse/ZOOKEEPER-1069 ported back to
> the 3.3 branch, but the tests written for these fixes won't port
> because the test case was never checked in to that branch. What should
> I do here? Check in the fixes to the branch with no test? I'm not
> entirely crazy about that option but it seems like that has been done
> before (the test in question was originally created for a different
> bug fix that was ported without tests to 3.3 from trunk).
> I'm ok to do that for these fixes but going forward do we really want
> to be pushing fixes in anywhere without tests? I'd hate to have a
> "fix" that doesn't work and we don't catch due to this.
>
> Thanks,
> C
>

The base test case this is extending, ZkTestCase, was never ported, so
I think it would be a bit more than just copying the test over. I
think backporting ZkTestCase isn't that big a deal, but then we also
have the whole slf4j upgrade that's going to start biting us over many
of the back ports we try to do. I could just take out the logging in
the test so I'll do that if we think that's the way to go.

On Sun, May 29, 2011 at 11:31 PM, Benjamin Reed <[EMAIL PROTECTED]> wrote:
> yeah, branches are a pain to maintain. we have done both ways: checked
> in the patch without the test and also backported the test. the later
> is the preferable one in my opinion. the test case is a whole new
> class right? can we just copy the class from trunk?
>
> ben
>
> On Sat, May 28, 2011 at 11:22 AM, Camille Fournier <[EMAIL PROTECTED]> wrote:
>> Hey guys,
>>
>> I'm trying to get the bug fixes for
>> https://issues.apache.org/jira/browse/ZOOKEEPER-1046 and
>> https://issues.apache.org/jira/browse/ZOOKEEPER-1069 ported back to
>> the 3.3 branch, but the tests written for these fixes won't port
>> because the test case was never checked in to that branch. What should
>> I do here? Check in the fixes to the branch with no test? I'm not
>> entirely crazy about that option but it seems like that has been done
>> before (the test in question was originally created for a different
>> bug fix that was ported without tests to 3.3 from trunk).
>> I'm ok to do that for these fixes but going forward do we really want
>> to be pushing fixes in anywhere without tests? I'd hate to have a
>> "fix" that doesn't work and we don't catch due to this.
>>
>> Thanks,
>> C
>>
>

yeah, this is why we skip the tests sometimes :)

your idea of changing the logging in the test does seem the best way to go.

thanx for sticking with this.

ben

On Mon, May 30, 2011 at 9:32 AM, Camille Fournier <[EMAIL PROTECTED]> wrote:
> The base test case this is extending, ZkTestCase, was never ported, so
> I think it would be a bit more than just copying the test over. I
> think backporting ZkTestCase isn't that big a deal, but then we also
> have the whole slf4j upgrade that's going to start biting us over many
> of the back ports we try to do. I could just take out the logging in
> the test so I'll do that if we think that's the way to go.
>
> On Sun, May 29, 2011 at 11:31 PM, Benjamin Reed <[EMAIL PROTECTED]> wrote:
>> yeah, branches are a pain to maintain. we have done both ways: checked
>> in the patch without the test and also backported the test. the later
>> is the preferable one in my opinion. the test case is a whole new
>> class right? can we just copy the class from trunk?
>>
>> ben
>>
>> On Sat, May 28, 2011 at 11:22 AM, Camille Fournier <[EMAIL PROTECTED]> wrote:
>>> Hey guys,
>>>
>>> I'm trying to get the bug fixes for
>>> https://issues.apache.org/jira/browse/ZOOKEEPER-1046 and
>>> https://issues.apache.org/jira/browse/ZOOKEEPER-1069 ported back to
>>> the 3.3 branch, but the tests written for these fixes won't port
>>> because the test case was never checked in to that branch. What should
>>> I do here? Check in the fixes to the branch with no test? I'm not
>>> entirely crazy about that option but it seems like that has been done
>>> before (the test in question was originally created for a different
>>> bug fix that was ported without tests to 3.3 from trunk).
>>> I'm ok to do that for these fixes but going forward do we really want
>>> to be pushing fixes in anywhere without tests? I'd hate to have a
>>> "fix" that doesn't work and we don't catch due to this.
>>>
>>> Thanks,
>>> C
>>>
>>
>

Hi Camille,

ZkTestCase really just introduces some improved logging/tracking
during the tests iirc. There is no need to backport it, just update
the test in question to extend TestCase (from junit) instead.

Look at the diffs btw 3.3 and trunk for something like:
src/java/test/org/apache/zookeeper/test/KeeperStateTest.java

It should be pretty straightforward. LMK if you have questions.

Patrick

On Mon, May 30, 2011 at 9:58 AM, Benjamin Reed <[EMAIL PROTECTED]> wrote:
> yeah, this is why we skip the tests sometimes :)
>
> your idea of changing the logging in the test does seem the best way to go.
>
> thanx for sticking with this.
>
> ben
>
> On Mon, May 30, 2011 at 9:32 AM, Camille Fournier <[EMAIL PROTECTED]> wrote:
>> The base test case this is extending, ZkTestCase, was never ported, so
>> I think it would be a bit more than just copying the test over. I
>> think backporting ZkTestCase isn't that big a deal, but then we also
>> have the whole slf4j upgrade that's going to start biting us over many
>> of the back ports we try to do. I could just take out the logging in
>> the test so I'll do that if we think that's the way to go.
>>
>> On Sun, May 29, 2011 at 11:31 PM, Benjamin Reed <[EMAIL PROTECTED]> wrote:
>>> yeah, branches are a pain to maintain. we have done both ways: checked
>>> in the patch without the test and also backported the test. the later
>>> is the preferable one in my opinion. the test case is a whole new
>>> class right? can we just copy the class from trunk?
>>>
>>> ben
>>>
>>> On Sat, May 28, 2011 at 11:22 AM, Camille Fournier <[EMAIL PROTECTED]> wrote:
>>>> Hey guys,
>>>>
>>>> I'm trying to get the bug fixes for
>>>> https://issues.apache.org/jira/browse/ZOOKEEPER-1046 and
>>>> https://issues.apache.org/jira/browse/ZOOKEEPER-1069 ported back to
>>>> the 3.3 branch, but the tests written for these fixes won't port
>>>> because the test case was never checked in to that branch. What should
>>>> I do here? Check in the fixes to the branch with no test? I'm not
>>>> entirely crazy about that option but it seems like that has been done
>>>> before (the test in question was originally created for a different
>>>> bug fix that was ported without tests to 3.3 from trunk).
>>>> I'm ok to do that for these fixes but going forward do we really want
>>>> to be pushing fixes in anywhere without tests? I'd hate to have a
>>>> "fix" that doesn't work and we don't catch due to this.
>>>>
>>>> Thanks,
>>>> C
>>>>
>>>
>>
>

Hi Camille,

I think I should have cleaned up the test before submitting it. Let me know
if you need some help with the merge.

-Vishal

On Mon, May 30, 2011 at 8:13 PM, Patrick Hunt <[EMAIL PROTECTED]> wrote:

> Hi Camille,
>
> ZkTestCase really just introduces some improved logging/tracking
> during the tests iirc. There is no need to backport it, just update
> the test in question to extend TestCase (from junit) instead.
>
> Look at the diffs btw 3.3 and trunk for something like:
> src/java/test/org/apache/zookeeper/test/KeeperStateTest.java
>
> It should be pretty straightforward. LMK if you have questions.
>
> Patrick
>
> On Mon, May 30, 2011 at 9:58 AM, Benjamin Reed <[EMAIL PROTECTED]> wrote:
> > yeah, this is why we skip the tests sometimes :)
> >
> > your idea of changing the logging in the test does seem the best way to
> go.
> >
> > thanx for sticking with this.
> >
> > ben
> >
> > On Mon, May 30, 2011 at 9:32 AM, Camille Fournier <[EMAIL PROTECTED]>
> wrote:
> >> The base test case this is extending, ZkTestCase, was never ported, so
> >> I think it would be a bit more than just copying the test over. I
> >> think backporting ZkTestCase isn't that big a deal, but then we also
> >> have the whole slf4j upgrade that's going to start biting us over many
> >> of the back ports we try to do. I could just take out the logging in
> >> the test so I'll do that if we think that's the way to go.
> >>
> >> On Sun, May 29, 2011 at 11:31 PM, Benjamin Reed <[EMAIL PROTECTED]>
> wrote:
> >>> yeah, branches are a pain to maintain. we have done both ways: checked
> >>> in the patch without the test and also backported the test. the later
> >>> is the preferable one in my opinion. the test case is a whole new
> >>> class right? can we just copy the class from trunk?
> >>>
> >>> ben
> >>>
> >>> On Sat, May 28, 2011 at 11:22 AM, Camille Fournier <[EMAIL PROTECTED]>
> wrote:
> >>>> Hey guys,
> >>>>
> >>>> I'm trying to get the bug fixes for
> >>>> https://issues.apache.org/jira/browse/ZOOKEEPER-1046 and
> >>>> https://issues.apache.org/jira/browse/ZOOKEEPER-1069 ported back to
> >>>> the 3.3 branch, but the tests written for these fixes won't port
> >>>> because the test case was never checked in to that branch. What should
> >>>> I do here? Check in the fixes to the branch with no test? I'm not
> >>>> entirely crazy about that option but it seems like that has been done
> >>>> before (the test in question was originally created for a different
> >>>> bug fix that was ported without tests to 3.3 from trunk).
> >>>> I'm ok to do that for these fixes but going forward do we really want
> >>>> to be pushing fixes in anywhere without tests? I'd hate to have a
> >>>> "fix" that doesn't work and we don't catch due to this.
> >>>>
> >>>> Thanks,
> >>>> C
> >>>>
> >>>
> >>
> >
>