I've finally pushed this change.
On 19 August 2016 at 15:13, Andrew Cagney <andrew.cag...@gmail.com> wrote: > Currently, when checking the results from a test, missing reference > output is ignored. For instance, the test: > > $ ls -1 testing/pluto/interop-ikev2-racoon-01-noconn/{,OUTPUT/} > testing/pluto/interop-ikev2-racoon-01-noconn/: > description.txt > eastinit.sh > east-racoon > final.sh > OUTPUT > testparams.sh > west.conf > westinit.sh > westrun.sh > > testing/pluto/interop-ikev2-racoon-01-noconn/OUTPUT/: > debug.log > east.console.txt > east.console.verbose.txt > pluto.east.log > RESULT > west.console.txt > west.console.verbose.txt > > has no reference output at all, yet it is considered a "good" test and > passes. Perhaps the intent was to detect a core dump, but that isn't > robust. > > I believe the rationale for this behaviour was that, in some cases, > getting output to match is "just too hard" (I can kind of believe it); > however I think a better way to deal with it would be to explicitly > sanitize away the output that should be ignored. For instance, by > putting: > > : =================== end ================== > > in the relevant *init.sh file. > > Ignoring "nic", there are ~18 "good" tests that have this problem; I > think they be changed to "wip". > In addition there are ~165 "good" tests that use "nic" (nic's output > is never checked!)', I think we can brute force them to just ignore > "nic"s output > > Andrew > > PS: Why? I've tests that need to check "nic"s output, and maybe doing > it isn't exactly robust :-/ _______________________________________________ Swan-dev mailing list Swan-dev@lists.libreswan.org https://lists.libreswan.org/mailman/listinfo/swan-dev