2010/5/26 Mark Hindess <mark.hind...@googlemail.com> > > In message <aanlktiklmus8j_qixdjbsxzwz9cyj9sdtuofmdqb6...@mail.gmail.com>, > "Jimmy,Jing Lv" writes: > > > > 2010/5/25 sebb <seb...@gmail.com> > > > > > On 25/05/2010, Jimmy,Jing Lv <firep...@gmail.com> wrote: > > > > > > > > [snip] > > > > > > > > For the first milestone of Harmony-Select, I suggest we'd > > > > have following criterion: > > > > 1. build pass on both windows/linux platforms; > > > > 2. pass all Harmony unit testcases in the selected module on > > > > these platforms; > > > > 3. Choose Apache Hadoop as a real Application test, pass its > > > > common test on these platforms for the first milestone. > > > > > > > > Wait for any kindly suggestions, opinions and comments. > > > > > > May suggest using some of the Commons components as test cases? > > > > > > For example, Commons Lang, Math, IO, Net? > > > > > > I know that Math has exposed bugs in Harmony in the past. > > Indeed. I found the commons-collections tests to be particularly > helpful and httpclient tests exposed a regex bug just last week when I > tried the recent release on our milestone candidates. > > > > These are relatively small codebases, so it should be a lot easier to > > > debug any failures. > > > > Hi, > > > > That's good suggestions, let's discuss the modules selection openly > > according to the application requirements. > > A silly question, as I haven't touched Commons yet, what is the > > advantage in Commons Lang, Math, IO/Net than J2SE components? > > As Tim says, I think Sebb was suggesting looking at the building blocks > of major projects and running their tests as a way to reach your goal > of running real applications. > > For instance, looking at: > > http://svn.apache.org/repos/asf/hadoop/common/trunk/ivy.xml > > I see that hadoop has pre-reqs for various use cases on[0]: > > commons-logging > lucene > commons-httpclient > commons-cli > commons-codec > commons-net > tomcat > commons-el > mina > ftpserver > commons-lang > commons-collections > etc. > > Rather than jumping straight to hadoop testing, it would be sensible to > try the tests from the pre-req projects first[1]. Problems found with > these tests would likely be easier to track down and fix than problems > in a large application. > > Thanks Tim and Mark (and of course Sebb!) This is very helpful for our goal, let's add them in select work items.
> Regards, > Mark. > > [0] I've probably missed some of the Apache projects and no doubt some of > the non-apache projects have great tests too. > > [1] Of course, if your goal is just Hadoop you'll undoubtedly find/fix > bugs that are out-of-scope for Hadoop successfully running but I don't > think running Hadoop is our only goal for Select. > > > -- Best Regards! Jimmy, Jing Lv China Software Development Lab, IBM
