[tests] increase timeouts in sendheaders test #10365

Increases short timeouts in the sendheaders.py test, which were causing intermittent failure when there was disk i/o contention.

We believe the root cause of this was that bitcoind log writes were excessively slow (up to several seconds) when another test was running in parallel and shutdown the node. The wallet flush on shutdown can block disk i/o for several seconds.

Longer term, it may be worth thinking about changing bitcoind LogPrints to write to a circular buffer and have a background thread write from there to disk. For now, let's just make this test less flakey. @sdaftuar

nitting: could you set a constant to that value instead of hunting and pecking for the number in case it changes again?

@instagibbs good point. wait_until() currently defaults to having no timeout. I can't see how that's useful - we'd want any timed predicate to eventually time out and fail the test. I propose we change wait_until() to have a 60 second timeout by default, and then remove the timeout parameter from almost all calls to wait_until() Definition is here: https://github.com/bitcoin/bitcoin/blob/master/test/functional/test_framework/mininode.py#L1360 @sdaftuar @MarcoFalke - thoughts?

@jnewbery Makes sense to me; I actually thought we already had a 60 second default and was going to suggest just using that when I saw @instagibbs' comment!

Updated wait_until() to have a default timeout of 60 seconds and removes the explicit timeouts from sendheaders.py

utACK https://github.com/bitcoin/bitcoin/pull/10365/commits/56befa03ac559ddab95eac4c2167b586dc0d01b9

utACK 56befa0

utACK 56befa03ac559ddab95eac4c2167b586dc0d01b9 though it might be nice to default the timeout in wait_for_* to float('inf') instead of 60 as well.

it might be nice to default the timeout in wait_for_* to float('inf') instead of 60

Which wait_for_ functions are you referring to? The ones in sendheader.py or in mininode.py? Why is changing the default to float('inf') better?