You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@river.apache.org by Zsolt Kúti <la...@gmail.com> on 2020/08/10 11:14:52 UTC
Re: Why jtreg
Hi guys,
Based on the SVN and Dennis' git repos, I now have a local git repo that
contains a gradlefied new test module. It accomplishes little for the
time being:
- follows River's gradle project format
- rectified (removed unused imports, corrected some package definitions
etc.) that prevented compiling
- can now be compiled
I planned to create at least one working Spock unit test to see how it
works out, but unfortunately was overwhelmed with other duties.
I thought that moving unit tests to their respective modules may come in a
later phase, or as soon as a new test is created.
Now, before leaving for long holidays it may be the time to merge this work
to somewhere.
Should I make my own github repo and put these changes there? I do not know
how pull request on Dennis' would work with large number of files/changes
(~ 5000).
Ideas?
Zsolt
On Tue, Jul 14, 2020 at 12:23 AM Peter Firmstone <
peter.firmstone@zeus.net.au> wrote:
> Why not indeed :)
>
> I'd be in favour of moving "unit tests" out of the jtreg and qa tests
> suites into junit first . I'd suggesting leaving the jtreg bug
> regression tests where they are for now at least, as far more work is
> required and they may not all be relevant; we no longer have bug
> descriptions or information on these bug's, as Oracle has removed them
> from the Sun bug database. I have looked into the undocumented
> regression tests previously, some are very difficult to figure out. I
> have made some attempts at recovering information on older bugs for
> these regression tests from release notes of earlier versions of Jini,
> but haven't had much luck, requests for information from Oracle go
> unanswered.
>
> The tests can be broken down into:
>
> 1. Unit tests - simple tests that don't need more than one running jvm,
> eg a lookup service, or activation and don't require a
> SecurityManager (maybe junit is ok with a security manager? Just
> need appropriate policy files?).
> 2. Integration tests - requires multiple jvm's, eg testing network
> functionality (typically in the qa suite).
> 3. Bug regression tests - testing a known, or often in our case, an
> undocumented bug.
>
> River never received an upload of the Sun bug database relating to Jini,
> Oracle has long since made it inaccessible. Many of the bug regression
> tests in jtreg lack documentation, I guess there might be some
> information in the Jini users mail list archives.
>
> I'd suggest grabbing the low hanging fruit first.
>
> Cheers,
>
> Peter.
>
> On 7/14/2020 6:58 AM, Dennis Reedy wrote:
> > As the title says, why use jtreg? We have modern test frameworks (Junit,
> > Spock, etc...). Asd we move forward with River, why not migrate tests to
> > use these?
> >
> > Regards
> >
> > Dennis Reedy
> >
>
Re: Why jtreg
Posted by Dennis Reedy <de...@gmail.com>.
Hi Zsolt,
Create a pull request and I’ll merge it in
Thanks
Dennis
> On Aug 10, 2020, at 7:15 AM, Zsolt Kúti <la...@gmail.com> wrote:
>
>
> Hi guys,
>
> Based on the SVN and Dennis' git repos, I now have a local git repo that contains a gradlefied new test module. It accomplishes little for the time being:
> - follows River's gradle project format
> - rectified (removed unused imports, corrected some package definitions etc.) that prevented compiling
> - can now be compiled
>
> I planned to create at least one working Spock unit test to see how it works out, but unfortunately was overwhelmed with other duties.
> I thought that moving unit tests to their respective modules may come in a later phase, or as soon as a new test is created.
>
> Now, before leaving for long holidays it may be the time to merge this work to somewhere.
> Should I make my own github repo and put these changes there? I do not know how pull request on Dennis' would work with large number of files/changes (~ 5000).
>
> Ideas?
>
> Zsolt
>
>> On Tue, Jul 14, 2020 at 12:23 AM Peter Firmstone <pe...@zeus.net.au> wrote:
>> Why not indeed :)
>>
>> I'd be in favour of moving "unit tests" out of the jtreg and qa tests
>> suites into junit first . I'd suggesting leaving the jtreg bug
>> regression tests where they are for now at least, as far more work is
>> required and they may not all be relevant; we no longer have bug
>> descriptions or information on these bug's, as Oracle has removed them
>> from the Sun bug database. I have looked into the undocumented
>> regression tests previously, some are very difficult to figure out. I
>> have made some attempts at recovering information on older bugs for
>> these regression tests from release notes of earlier versions of Jini,
>> but haven't had much luck, requests for information from Oracle go
>> unanswered.
>>
>> The tests can be broken down into:
>>
>> 1. Unit tests - simple tests that don't need more than one running jvm,
>> eg a lookup service, or activation and don't require a
>> SecurityManager (maybe junit is ok with a security manager? Just
>> need appropriate policy files?).
>> 2. Integration tests - requires multiple jvm's, eg testing network
>> functionality (typically in the qa suite).
>> 3. Bug regression tests - testing a known, or often in our case, an
>> undocumented bug.
>>
>> River never received an upload of the Sun bug database relating to Jini,
>> Oracle has long since made it inaccessible. Many of the bug regression
>> tests in jtreg lack documentation, I guess there might be some
>> information in the Jini users mail list archives.
>>
>> I'd suggest grabbing the low hanging fruit first.
>>
>> Cheers,
>>
>> Peter.
>>
>> On 7/14/2020 6:58 AM, Dennis Reedy wrote:
>> > As the title says, why use jtreg? We have modern test frameworks (Junit,
>> > Spock, etc...). Asd we move forward with River, why not migrate tests to
>> > use these?
>> >
>> > Regards
>> >
>> > Dennis Reedy
>> >