Greg Smith

for issue https://github.com/libuv/libuv/issues/3487

Sure, it'll take a day or so

added additional tests to existing file. Linux behavior and windows behavior match (with respect to delay)

Added requested changes. I'm torn on not setting TCP_KEEPINTVL and TCP_KEEPCNT to the windows ( and documentation ) values in unix when delay == 0 ( I painted myself into...

I don't have access to any other platforms, so I'm not sure how to answer. I thought I saw an issue with keepalives on an IBM OS during my original...

Note sure how downstream use ( Ie Node ) binds libuv behavior but Node expects: Setting 0 for initialDelay will leave the value unchanged from the default (or previous) setting...

I think in your example the second call would fall under "Setting 0 for initialDelay will leave the value unchanged from the ...previous.. setting" clause in the doc. I would...

tested, delay stays at 60. I believe that is acceptable. ``` uv_tcp_keepalive(handle, 1, 60); uv_tcp_keepalive(handle, 0, 0); /* disable keepalives */ uv_tcp_keepalive(handle, 1, 0); /* re-enable keepalives...delay is still 60...

Linux has same behavior. If delay is 0 the system default is used ( unless previously set, then that previous value is used)

I guess whether the behavior is surprising or not, depends on why someone would try to set the delay multiple times or disable/enable keepalives after enabling them. I can't that...