-=| gregor herrmann, Tue, Dec 29, 2009 at 11:24:08AM +0100 |=- > On Mon, 21 Dec 2009 18:11:59 +0200, Damyan Ivanov wrote: > > The failing test file has two tests. Both check whether some > > resource consumption is approximately the same as what times() > > reports. There is a tolerance (20% for user time and 40% for > > system time) and the tests fail if the measured resource > > consumprion diverges by aa great factor. > > The tests point to INSTALL which has an interested paragraph > concerning this specific test: > > #v+ > NOTE 1: The test t/getrusage.t is known to be tricky and to fail once > in a while. This is because the results depend on timing issues and > the activity of the operating system, it is moderately icky to > guestimate such tolerances that leave some slack while still proving > anything. Please retry 'make test' for a couple times, "eventually" > the test should pass. If even that fails, please see that the $debug > in the script is 1 and then run the test separately: > > perl -Mblib t/getrusage.t > > or (if the above does not work) > > perl -Iblib/arch -Iblib/lib t/getrusage.t > > and send the results to me. Ditto for the other tests: running them > separately with the $debug on is useful for, well, debugging. > #v-
Nice catch! We could loop tests several times until either a) they pass; or b) N loops were done and then try one final loop with debugging on. At least this can help upstream help us :) -- dam
signature.asc
Description: Digital signature
