You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@edgent.apache.org by Dale LaBossiere <dm...@gmail.com> on 2017/09/20 21:14:38 UTC
maven-cvt: console regressions
> On Sep 20, 2017, at 1:32 PM, Christofer Dutz <ch...@c-ware.de> wrote:
> ...
> I’m sure that is the reason for this. I did investigate that problem a wile yesterday and I found out that the MBean the ConsoleMetricsServlet->MetricsUtil (line 105) is using to generate the output returns “long” instead of “counter” … but I have to admit that I have no Idea why.
>
> Will investigate tomorrow ;-)
I found another regression. The hover on a union oplet incorrectly reports the output tuple count - it’s not the sum.
It’s demonstrable with the ConsoleWaterDetector sample (suspect caused by the same problem causing the other regression):
cd samples
mvn package
cd console
./run-samples.sh ConsoleWaterDetector
I did a review of the cleanup changes that had been made to MetricsUtil.java and found one problem though fixing it didn’t address either of these regressions. I’m delivering more cleanup of it.
— Dale
Re: maven-cvt: console regressions
Posted by Christofer Dutz <ch...@c-ware.de>.
One of my easiest exercises … but merge first, release second and then: let the fun begin ;-)
Chris
Am 25.09.17, 14:21 schrieb "Dale LaBossiere" <dm...@gmail.com>:
It would be fabulous to have some automated testing for the console (post getting a release out :-)
— Dale
> On Sep 25, 2017, at 8:15 AM, Christofer Dutz <ch...@c-ware.de> wrote:
> ...
> But it does show that it might be good to start doing, what I had introduced for the Flex project. As soon as we merge things back, I think it would be a good idea to start run something that utilizes the entire stack and verified via Selenium that the console output is as desired.
Re: maven-cvt: console regressions
Posted by Dale LaBossiere <dm...@gmail.com>.
It would be fabulous to have some automated testing for the console (post getting a release out :-)
— Dale
> On Sep 25, 2017, at 8:15 AM, Christofer Dutz <ch...@c-ware.de> wrote:
> ...
> But it does show that it might be good to start doing, what I had introduced for the Flex project. As soon as we merge things back, I think it would be a good idea to start run something that utilizes the entire stack and verified via Selenium that the console output is as desired.
Re: maven-cvt: console regressions
Posted by Christofer Dutz <ch...@c-ware.de>.
Hi Dale,
Argh! … ok that seemed to be so obviously wrong that my brain didn’t even seem to remember “fixing” that.
Maybe I shouldn’t delegate fixing things to my co-processor ;-)
But it does show that it might be good to start doing, what I had introduced for the Flex project. As soon as we merge things back, I think it would be a good idea to start run something that utilizes the entire stack and verified via Selenium that the console output is as desired.
Chris
Am 25.09.17, 13:55 schrieb "Dale LaBossiere" <dm...@gmail.com>:
The problem was introduced here: https://github.com/apache/incubator-edgent/pull/309/commits/78a3e6fdc935993a5c0445e16e732ae25bbc6b52
obj.opId was set correctly on line 52 (from op.opId). After the change the value (now from met.opId) I believe ended up being null which subsequently messed up other processing. Why the original no-op assignment was there… who knows? :-)
— Dale
> On Sep 21, 2017, at 7:52 AM, Christofer Dutz <ch...@c-ware.de> wrote:
>
> Hi Dale,
>
> could you please sum up what the problem was? I did see you simplifying some of the code, but couldn’t pin-point the individual change that fixed things.
>
> Chris
>
>
>
> Am 20.09.17, 23:56 schrieb "Dale LaBossiere" <dm...@gmail.com>:
>
> I just delivered the fix for the two regressions. See its comment for details.
> I think we can claim the console is operational again!
>
> — Dale
>
>> On Sep 20, 2017, at 5:14 PM, Dale LaBossiere <dm...@gmail.com> wrote:
>>
>>
>>> On Sep 20, 2017, at 1:32 PM, Christofer Dutz <ch...@c-ware.de> wrote:
>>> ...
>>> I’m sure that is the reason for this. I did investigate that problem a wile yesterday and I found out that the MBean the ConsoleMetricsServlet->MetricsUtil (line 105) is using to generate the output returns “long” instead of “counter” … but I have to admit that I have no Idea why.
>>>
>>> Will investigate tomorrow ;-)
>>
>> I found another regression. The hover on a union oplet incorrectly reports the output tuple count - it’s not the sum.
>> It’s demonstrable with the ConsoleWaterDetector sample (suspect caused by the same problem causing the other regression):
>>
>> cd samples
>> mvn package
>> cd console
>> ./run-samples.sh ConsoleWaterDetector
>>
>> I did a review of the cleanup changes that had been made to MetricsUtil.java and found one problem though fixing it didn’t address either of these regressions. I’m delivering more cleanup of it.
>>
>> — Dale
>
>
>
Re: maven-cvt: console regressions
Posted by Dale LaBossiere <dm...@gmail.com>.
The problem was introduced here: https://github.com/apache/incubator-edgent/pull/309/commits/78a3e6fdc935993a5c0445e16e732ae25bbc6b52
obj.opId was set correctly on line 52 (from op.opId). After the change the value (now from met.opId) I believe ended up being null which subsequently messed up other processing. Why the original no-op assignment was there… who knows? :-)
— Dale
> On Sep 21, 2017, at 7:52 AM, Christofer Dutz <ch...@c-ware.de> wrote:
>
> Hi Dale,
>
> could you please sum up what the problem was? I did see you simplifying some of the code, but couldn’t pin-point the individual change that fixed things.
>
> Chris
>
>
>
> Am 20.09.17, 23:56 schrieb "Dale LaBossiere" <dm...@gmail.com>:
>
> I just delivered the fix for the two regressions. See its comment for details.
> I think we can claim the console is operational again!
>
> — Dale
>
>> On Sep 20, 2017, at 5:14 PM, Dale LaBossiere <dm...@gmail.com> wrote:
>>
>>
>>> On Sep 20, 2017, at 1:32 PM, Christofer Dutz <ch...@c-ware.de> wrote:
>>> ...
>>> I’m sure that is the reason for this. I did investigate that problem a wile yesterday and I found out that the MBean the ConsoleMetricsServlet->MetricsUtil (line 105) is using to generate the output returns “long” instead of “counter” … but I have to admit that I have no Idea why.
>>>
>>> Will investigate tomorrow ;-)
>>
>> I found another regression. The hover on a union oplet incorrectly reports the output tuple count - it’s not the sum.
>> It’s demonstrable with the ConsoleWaterDetector sample (suspect caused by the same problem causing the other regression):
>>
>> cd samples
>> mvn package
>> cd console
>> ./run-samples.sh ConsoleWaterDetector
>>
>> I did a review of the cleanup changes that had been made to MetricsUtil.java and found one problem though fixing it didn’t address either of these regressions. I’m delivering more cleanup of it.
>>
>> — Dale
>
>
>
Re: maven-cvt: console regressions
Posted by Christofer Dutz <ch...@c-ware.de>.
Hi Dale,
could you please sum up what the problem was? I did see you simplifying some of the code, but couldn’t pin-point the individual change that fixed things.
Chris
Am 20.09.17, 23:56 schrieb "Dale LaBossiere" <dm...@gmail.com>:
I just delivered the fix for the two regressions. See its comment for details.
I think we can claim the console is operational again!
— Dale
> On Sep 20, 2017, at 5:14 PM, Dale LaBossiere <dm...@gmail.com> wrote:
>
>
>> On Sep 20, 2017, at 1:32 PM, Christofer Dutz <ch...@c-ware.de> wrote:
>> ...
>> I’m sure that is the reason for this. I did investigate that problem a wile yesterday and I found out that the MBean the ConsoleMetricsServlet->MetricsUtil (line 105) is using to generate the output returns “long” instead of “counter” … but I have to admit that I have no Idea why.
>>
>> Will investigate tomorrow ;-)
>
> I found another regression. The hover on a union oplet incorrectly reports the output tuple count - it’s not the sum.
> It’s demonstrable with the ConsoleWaterDetector sample (suspect caused by the same problem causing the other regression):
>
> cd samples
> mvn package
> cd console
> ./run-samples.sh ConsoleWaterDetector
>
> I did a review of the cleanup changes that had been made to MetricsUtil.java and found one problem though fixing it didn’t address either of these regressions. I’m delivering more cleanup of it.
>
> — Dale
Re: maven-cvt: console regressions
Posted by Dale LaBossiere <dm...@gmail.com>.
I just delivered the fix for the two regressions. See its comment for details.
I think we can claim the console is operational again!
— Dale
> On Sep 20, 2017, at 5:14 PM, Dale LaBossiere <dm...@gmail.com> wrote:
>
>
>> On Sep 20, 2017, at 1:32 PM, Christofer Dutz <ch...@c-ware.de> wrote:
>> ...
>> I’m sure that is the reason for this. I did investigate that problem a wile yesterday and I found out that the MBean the ConsoleMetricsServlet->MetricsUtil (line 105) is using to generate the output returns “long” instead of “counter” … but I have to admit that I have no Idea why.
>>
>> Will investigate tomorrow ;-)
>
> I found another regression. The hover on a union oplet incorrectly reports the output tuple count - it’s not the sum.
> It’s demonstrable with the ConsoleWaterDetector sample (suspect caused by the same problem causing the other regression):
>
> cd samples
> mvn package
> cd console
> ./run-samples.sh ConsoleWaterDetector
>
> I did a review of the cleanup changes that had been made to MetricsUtil.java and found one problem though fixing it didn’t address either of these regressions. I’m delivering more cleanup of it.
>
> — Dale