On 9/11/07, Sean Qiu <[EMAIL PROTECTED]> wrote: > Or we can add a new adapor for the emma task, called "emma" as a example. > The benifit is we needn't modify the main script. > > 1. emma/parameter.xml > make it depends "classlib" and "drlvm", get the latest built jre for > instrument. > tell the instrumented jre location as a shared parameter > 2. emma/adaptor.xml > instrument the jre > 3. setup wanted test to get coverage report. > set wanted test suites' depends parameter to "emma", and take the > instrumented jre as the test jre. > 4. run the test to get coverage report > > But i think there is a little violation against the bti adaptor's > abstraction, so i prefer the previous one. > > Any suggestion or commets?
Besides, I have put a kernel classes list on harmony wiki.[1] You may try it, Sean.:) It is generated by a simulation of emma's gathering coverage data. If any future modification in harmony's code influences the emma depending classes, a solution to generate kernel classes list is also included in the wiki page[2]. Good luck! [1]http://wiki.apache.org/harmony-data/attachments/coverageEMMA/attachments/kernel.classes.list [2]http://wiki.apache.org/harmony/coverageEMMA > > 2007/9/10, Sean Qiu <[EMAIL PROTECTED]>: > > > > Hi, guys > > > > Shall we integrate the emma into our BTI 2.0 to get the test coverage > > report? > > AFAIK, Robert has spent a number of time on using our unit test to emma > > coverage report. > > Maybe we can generate the coverage from the BTI test besides our unit > > test. > > > > IMHO, we can maintain a individual target as the "run-cc" target in > > script/main.xml, like "run-coverage" or something else. > > It will set up its requisite like an instrumented jre to run the test. > > Before running the test, we need to add <jvmarg > > value="-Xbootclasspath/p:${instrumented-classlib}"> for each adaptor's > > tested jvm task. > > > > Finally, the "run-coverage" command call each adaptor as normal except > > assigning the ${instrumented-classlib} to the instrumented classlib jars. > > The generated report can be placed to build/coverage-report or some more > > proper places. > > > > I think this approach can extend the BTI 2.0 without > > too many modifications. > > Are there any comments about this? Or any other approach? Any > > suggestion is welcomed. > > > > -- > > Sean Qiu > > China Software Development Lab, IBM > > > > > > -- > Sean Qiu > China Software Development Lab, IBM > -- Leo Li China Software Development Lab, IBM
