More japitools status

[Stuart Ballard]

Stuart Ballard wrote:
> For now, if people here don't object, I'll post occasional status 
> reports on the Classpath list which would otherwise have gone to a 
> japitools list.

Please do let me know if you'd prefer me not to post this stuff here.

> I'm still working on false-positive elimination from the results of 
> earlier JDKs. I have a pretty good handle on exactly what needs to 
> change where - it's a surprisingly deep and subtle change because if 
> you're not careful it breaks a lot of assumptions about things like 
> percentages adding up to 100% (what kind of moron would assume that, 
> honestly? ;) )

Good progress on this front. I have a version of japicompat that tracks 
which errors are due to compatibility with the latest JDK instead of the 
one being tested, skips reporting those errors, and provides data to the 
output filter (japiotext or japiohtml) to allow them to calculate the 
percentages accordingly.

I also have a version of japiohtml that understands the new data and 
makes an attempt to present it in a reasonable way. I'm currently 
tracking down issues in its percentage calculation and bar graph 
generation - in various places, 1px-wide colored bars show up in colors 
where the actual percentage is supposed to be zero, and one memorable 
report showed java.io as having "-0.3% bad". When I'm confident in the 
results, I'll switch to using this version to generate the "nightly" 
results.

Even so, my testing so far suggests that this drastically reduces the 
number of errors reported against earlier JDK versions, to the point 
where "zero errors against 1.0 and 1.1 except as necessary for 1.4 
compatibility" looks like it might be an easily achievable goal.

Be aware that the percentage scores from the new version are not 
directly comparable to the scores from the old, due to changes in the 
way they are calculated. Naturally it's impossible to give a 100% 
accurate percentage of something as vaguely-defined as "API coverage" so 
whatever approach I take will have some kinds of systemic errors. The 
"truth", inasmuch as it exists, lies somewhere between the old and new 
techniques, although I think it's closer to the new.

Stuart.

--
Stuart Ballard, Senior Web Developer
NetReach, Inc.
(215) 283-2300, ext. 126
http://www.netreach.com/