From: David A. M. <dm...@re...> - 2003-05-07 17:17:38
|
Manoj Iyer wrote: > > > > > > Like others responded select the testcase you want to execute and create > > > a custom command file and use the -f option to execute your custom list > > > of testcases. > > > > This looks like the best option for us to use, including only the > > testcases which are applicable to the system being tested, which brings > > me to my next question: Are the system requirements for these test > > cases documented? > > The best documentation available is 'code'. . . But there's a lot of code. ;-) Seriously, the code can tell you what each test does, but does not necessarily reveal resource requirements. Such things often require empirical data based on previous test runs. > . . . There is a small > descriptions file that gives you some idea as to what these tests are > supposed to do. Well it will be a great thing to have if someone came up > with something like that.! :-) As we run the tests and determine resource requirements and limits, we can document these and add them back to the descriptions file, if that would be appropriate. If others do the same, we can eventually develop a rather comprehensive list. > > > I would like to run all tests possible on a given target system, but > > need a way to identify which are possible. This leads to my third > > question: Has there been any effort to create a "tiered" test > > configuration? We were considering a way to select specific tests based > > on the system's capabilities, i.e., amount of memory, disk capacity (if > > present), network connection present, etc. In other words, a way to > > create a custom command file based on the system configuration. Does > > this sound like something that would be useful to others? > > In the next release of LTP there will be a menu driven scheme to > selectively execute testcase, this menu is based on ncurses etc, so > might not work well in your case. > > One solution to this could be to autoconf LTP. I am looking into it... > But if you can come up with a scheme that will be cool! autoconf can gather a lot of information about the 'build' system, but we are using a cross compiler for these systems, which limits the information available about the 'target' system (cannot determine amount of memory, storage, if network is present, etc.). The menu driven system would probably be appropriate since it could be run on the build system to configure the tests to be run on the target. I'll get the next release of LTP and see how the menu system works. If possible, we'll look into expanding this to configure tests based on the target system configuration (some set of criteria). Thanks again, d.marlin |