On 5/7/07, Alexey Petrenko wrote:
I've counted japitool result for the "endorsed" packages as I promised before. RI5 vs. Harmony M1 result is 98.33%. Harmony M1 vs. RI5 result is 99.78%. The results looks pretty good :)
I had a quick look at generated reports and found that column 'minor' was not included. That makes statistics to look a bit better :-) As I see the column counts SerialVersionUID incompatibilities. Do you think they can be ignored? -Stepan.
The results have been published on the web [1], [2]. Number of discovered issues is fixed in svn. SY, Alexey [1]http://people.apache.org/~apetrenko/japitool/jdk5vsharmony5.javajavax.html [2]http://people.apache.org/~apetrenko/japitool/harmony5vsjdk5.javajavax.html 2007/4/26, Mikhail Loenko <[EMAIL PROTECTED]>: > 2007/4/26, Alexey Petrenko <[EMAIL PROTECTED]>: > > And we can do this ourselves... > > thanks for volunteering! > > Thanks, > Mikhail > > > > > 2007/4/26, Mikhail Loenko <[EMAIL PROTECTED]>: > > > What we need I think is completeness metrics. Since those differences > > > that are caused by newer specs are OK, they don't affect completeness. > > > > > > So the real API completeness might be higher than what we see now. > > > Having the real picture would help us to first stick those really unimplemented > > > classes remained and second better position our state > > > > > > So it's good to have both: compatibility that Stuart is currently measuring > > > and completeness to what we want to achive. > > > > > > Thanks, > > > Mikhail > > > > > > 2007/4/26, Tim Ellison <[EMAIL PROTECTED]>: > > > > Mikhail Loenko wrote: > > > > > well, let's resume discussion whether Harmony must contain implementation > > > > > of the endorsed specs of the same version as RI or may contain a newer > > > > > version > > > > > > > > > > Tim, could you please comment on that? > > > > > > > > Only to say that the SE spec allows for us to implement a later version > > > > of these endorsed external specifications; however, Stuart is measuring > > > > compatibility to Sun's implementation, not compliance. We should expect > > > > to see a difference to the extent that the Corba code is incompatible > > > > with previous versions. > > > > > > > > Regards, > > > > Tim
