On Wed, 4 Oct 2017, Andrew Cagney wrote:
On 4 October 2017 at 10:38, Paul Wouters <[email protected]> wrote:But for that same reason, adding retransmit-interval=2000 everywhere is not good because people will copy that from our examples. And Antony is right that using RETRANSMIT_INTERVAL_DEFAULT=2000 as compile time value would address that. I still don't like running/testing code that doesn't match real life though. But it's a weak argument and as Antony pointed out, if you set it in almost all cases anyway, it is practically the same thing. I think --impair-retransmits almost but not quite does what we want (suppress a re-transmit when testing that the first packet dtrt). It doesn't leak into the config file; but it does [correctly] log the impairment. --impair-logging-of-impaired-retransmits or perhaps --impair-do-not-even-schedule-retransmits :-)
The impair _prevents_ retransmits. For regular success test cases, if the first packet is lost in the test rig, we _want_ a retransmit. We just don't want it to be visible it happened. Paul
_______________________________________________ Swan-dev mailing list [email protected] https://lists.libreswan.org/mailman/listinfo/swan-dev
