Hi Jan, I've finished all the analysis on libti code coverage.
For those covered functions, I listed all the exit point not hitted in them. And in conclusion section, there are funtions, feature or function exit point which are important to be covered by test driver or test suite. Please let me know whether I missed anything important, or anything not important. A question raised during analyzing: when create a target, where there is an existing target with the same name, with the same type or not, how will TI action? So for all types of target, what will TI do. I tried a case that creating a dump volume with the same name as an existing zfs, it report as pass, but actually nothing happened, no dump volume created, I think it would be better to report as a failure. And also I think the test suite should take some effort to test on this situation. Thanks, Angela angela ??: > Hi Jan, > > I'm still analyzing on function exit point coverage. > As the C code and the disassemble code are not one to one corespondent, > so I wrote some debug msg code in libti to trace exit point, the results > will not exactly the same with Qtrace. > I'll update you once finished. > > About current report, it is based on result got from Qtrace, a code > coverage tool. > And currently Qtrace only support SPARC. > So for i386 part, I can manually do some simple trace using debug msg code. > Will implement them also. > > Cheers, > Angela > > jan damborsky ??: > >> Hi Angela, >> >> my apologies for the delay. >> Looking at the test results, they are definitely useful. >> It can be seen what parts of code are not tested by >> current version of test driver and also I have found some >> pieces in TI which might be probably completely removed >> (I will need to take closer look). >> When changes are to be done in test driver I will take into >> account these reports as well, so that better coverage is >> accomplished. >> If my understanding is correct, those kind of coverage >> analysis can be only done on Sparc at the time being ? >> >> Thank you very much for these information ! >> Jan >> >> >> On 03/19/09 07:33, angela wrote: >> >>> Hi Jan, >>> >>> Recently I used Qtrace to trace library libti and analyze code >>> coverage by >>> libti test suite. >>> >>> You may find the code coverage result: >>> http://opg-qe.central.sun.com/wiki/index.php/Install_libti#Qtrace_result_on_libti_V1.0 >>> >>> * analysis >>> >>> <http://greatwall.prc/%7Ehl150050/qtrace/results/libti_1.0/libti.analysis/analysis> >>> : code coverage analysis result >>> * summary >>> >>> <http://greatwall.prc/%7Ehl150050/qtrace/results/libti_1.0/libti.analysis/summary> >>> : summary provided by Qtrace >>> * libti details function >>> >>> <http://greatwall.prc/%7Ehl150050/qtrace/results/libti_1.0/libti.analysis/details_func_libti.so.1> >>> : libti function details provided by Qtrace >>> >>> In analysis >>> <http://greatwall.prc/%7Ehl150050/qtrace/results/libti_1.0/libti.analysis/analysis>, >>> it includes function list that was hit or not, reason why not hit. >>> It also inludes the reason which exit entry not covered for some >>> functions, and >>> I will try to implement this section for all hit functions later. >>> >>> If you think something wrong or not accurate, please let me know. >>> Hope this can be useful to you. >>> >>> Thanks, >>> Angela >>> > >
