From: Paul L. <pl...@li...> - 2003-10-27 18:05:47
|
On Mon, 2003-10-27 at 10:04, Aaron McSorley wrote: > The last time I checked the pass/fail results were not reliable when > running parallel with -x, some tests would get in the way of others. > Has this changed in the newer versions of LTP. It depends on what you call "results". :) If you are looking for correctness, it really only makes sense to run one copy most of the time. However if you are trying to do a lot of weird things, mixed load/random stress by running a lot of copies in parallel then the results you're looking for are more likely to come out of the serial console. For instance, I know a guy who recently used 256 copies of LTP running simultaneously to reproduce a defect that had previously been very elusive. -Paul Larson |