Add block=True back to get_msg()

[davidbrochart]

See #640 (comment)

[kevin-bates]

I was going to comment on whether or not we should deprecate the block parameter at this time (and remove it in 7.0) and/or document that timeout is only used when block == True (although the comment was the same previously). Kind of an odd API and agree that the timeout value serves that purpose just fine. We should probably default its value to 0 once block is removed (IMO).

[davidbrochart]

Ah, sorry for merging so quickly. Actually, now that we decided to go for a major release and it's still a pre-release, we could remove it.
But defaulting timeout to 0 means we just poll for a message, which is the opposite of what we have now by default (we wait until we have a message)?

[kevin-bates]

But defaulting timeout to 0 means we just poll for a message, which is the opposite of what we have now by default (we wait until we have a message)?

I see. I haven't looked at the socket poll() method, but now see what you're saying where None essentially means wait until we have a message. Yeah, that should probably remain the default then.

[kevin-bates]

@davidbrochart - was there still follow-up work for the 7.0 release on this? I was under the impression the block parameter would be removed but see it's still there. Is this just a matter of removing the parameter and leaving timeout defaulted to None? (Essentially make the change that was backed out of the 6.x releases?)

[davidbrochart]

I think we can get rid of the block parameter, since we can do anything with timeout only. I opened #671.