From: William J. H. <wj...@us...> - 2003-01-13 19:44:19
|
I can say from our perspective right now we only run the kernel section. Although it's not what I would prefer it's just the way things are currently. I hope the kernel development community becomes more consistent at checking their work with LTP (regression tests being a big part). That reduces our superficial efforts at running LTP as a glibc/kernel validation tool and we can start exploring the other testcases (network) which can get more interesting. Because as we all know just cause the network can send data doesn't mean it works if it consumes 80% of the CPU. Things like the network test tools, and other tools which have a dual purpose, like growfiles for example, are interesting. It not only tests parts of the syscall layers, but it also tests hardware and io and so forth. If combined with say a network IO tool you can really stress test that particular subsystem i a "real world" type of test. And maybe come up with a magic simulation # for pretenting to be a webserver. Again, once the kernel work is pretty much in effect I think there's a cause for branching out into more of a "system test tool". You can get into interesting things too like maybe being able to plug in the "contest" kernel benchmarking tool along with other tools say "bonnie" (which may be part of contest already now that I think of it) and whatever else and then LTP could be run as an overall system benchmark / stress tool. I don't know if I see the command testing section as a big priority. Simply because I feel that most cmds have really stagnated and the "does ar support the -v flag" test doesn't really tell you much about whether or not it's a real bug for a lot of effort just getting there. But that's just my reasoning. I really liked the idea that someone posted about trying to use automake's testing for ./configure scripts. There are also tools that just randomly pick options and feed in garbage as a means of checking for buffer overflows. More generic tools rather then a test case for each cmdline tool. Just picking say emacs or vi (way too many options / combinations), even say ping and traceroute (potentially too much setup), ls (how do you know the #'s it's returning are valid). Each of the examples illustrates one reason why it can be problematic to create a testcase for that cmdline tool. Now combine the possibilities of multiple issues for the same tool, along with an unfavorable multiplication rate (1 tool * ~10 options) seems to indicate to me the low yield for a given amount of effort. My vision at least. j Jay Huie wj...@us... zSeries Linux System Test Phone: 845-435-8164 Sent by: ltp...@li... To: Robert Williamson/Austin/IBM@IBMUS cc: ltp <ltp...@li...> Subject: Re: [LTP] Modularized gzip'd tarballs created.... Robbie, I think we should understand what percent of LTP network or Command section are used by those who use LTP regularly. It will be nice to get some feedback on this from those using LTP. If the testing community doesint care about the large tgz file or RPM they have to download, we dont have to take the trouble of modularizing LTP. Personally I think there are benifits in the long run by modularizing LTP. My thought is, the runall script should have options to run specific sections of LTP, also, change the name of that script to run-ltp.sh something like that. run-ltp.sh should be able to handle a scenario file based test execution, this will provide flexibility. If there is support/agreement on such issues in the LTP-community I can work on a script. Cheers Manjo ******************************************************************************* cognito ergo sum ******************************************************************************* On Mon, 13 Jan 2003, Robert Williamson wrote: > As promised, I created the gzip'd tarball versions of the RPMs created > Friday. The network and commands modules are just simply the directories, > that would need to be placed under /testcases when opened. This is all > being done in anticipation of the commands section getting bigger. I may > even toy around with modularizing the kernel section.....but that, if ever, > will come later. > > - Robbie > > Robert V. Williamson <ro...@us...> > Linux Test Project > IBM Linux Technology Center > Phone: (512) 838-9295 T/L: 678-9295 > Fax: (512) 838-4603 > http://ltp.sourceforge.net > ==================== > "Only two things are infinite, the universe and human stupidity, and I'm > not sure about the former." -Albert Einstein > > > > > ------------------------------------------------------- > This SF.NET email is sponsored by: FREE SSL Guide from Thawte > are you planning your Web Server Security? Click here to get a FREE > Thawte SSL guide and find the answers to all your SSL security issues. > http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0026en > _______________________________________________ > Ltp-list mailing list > Ltp...@li... > https://lists.sourceforge.net/lists/listinfo/ltp-list > ------------------------------------------------------- This SF.NET email is sponsored by: FREE SSL Guide from Thawte are you planning your Web Server Security? Click here to get a FREE Thawte SSL guide and find the answers to all your SSL security issues. http://ads.sourceforge.net/cgi-bin/redirect.pl?thaw0026en _______________________________________________ Ltp-list mailing list Ltp...@li... https://lists.sourceforge.net/lists/listinfo/ltp-list |