harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Alex Orlov" <amor...@gmail.com>
Subject Re: Regression tests (was: svn commit: r389475 - in /incubator/harmony/enhanced/classlib/trunk/modules/text/src: main/java/java/text/SimpleDateFormat.java test/java/tests/api/java/text/SimpleDateFormatTest.java)
Date Tue, 28 Mar 2006 13:38:18 GMT
On 3/28/06, Mark Hindess <mark.hindess@googlemail.com> wrote:
> I'm not convinced.
>
> Having the JIRA reference for non-contentious tests - such as an API
> test that passes on the reference implementation - seems a little
> redundant since no one will ever bother to look it up.  If a test was
> contentious then I'd rather see a brief summary/explanation with the
> code (perhaps referencing the JIRA) than to have only the JIRA
> reference.  Having just the JIRA reference seems like a poor shortcut for
> the important cases and needless clutter for the rest.
>
> If we really need the information later it is in the svn log anyway?
> For instance, I just did:
>
> bash$ svn blame \
> modules/text/src/test/java/tests/api/java/text/SimpleDateFormatTest.java| \
> grep test_equals_afterFormat
> 389475   tellison     public void test_equals_afterFormat() {
> bash$ svn log -r 389475
> ------------------------------------------------------------------------
> r389475 | tellison | 2006-03-28 11:56:21 +0100 (Tue, 28 Mar 2006) | 2 lines
>
> Fix for HARMONY-209 (java.text.SimpleDateFormat.equals() behavior is
> not compatible with RI)
>
> ------------------------------------------------------------------------
>
> Granted it gets a little more difficult over time, but the svn log
> doesn't go missing or get deleted accidentally, forgottent, etc.  I
> prefer to let
> the version control system store this kind of metadata (since that is
> what it is intended for) rather than cluttering the code with it.

When the tests are run and some of them will fail it will be quite
convenient to look in the test sources and see its Java code and
whether it contains reference to the corresponding JIRA issue.
Otherwise for each unit test failure someone will have to look in CVS
log even if this is not regression test indeed.

However I agree it makes perfect sense to add brief
summary/explanation in addition to JIRA code.

--
Alex Orlov,
Intel Middleware Products Division.

>
>
> Regards,
> Mark.
>
> On 3/28/06, Tim Ellison <t.p.ellison@gmail.com> wrote:
> > Yep, I agree this is a good convention. I missed it here but will go and
> > fix it.
> >
> > Thanks
> > Tim
> >
> > Stepan Mishura wrote:
> > > Hi,
> > >
> > > We agreed some time ago to mark regression tests with comments that includes
> > > a reference to JIRA issue.
> > >  There were no objections and I believe this is the good rule so I'd like to
> > > ask everybody to follow it.
> > > I've update testing wiki page to fix this agreement (see
> > > http://wiki.apache.org/harmony/Testing_Convention)
> > >
> > > Thanks,
> > > Stepan.
>
> --
> Mark Hindess <mark.hindess@googlemail.com>
> IBM Java Technology Centre, UK.
>

Mime
View raw message