[PATCH remotefilelog-ext] Reuse ssh connection across miss fetches

[Augie Fackler]

> On Dec 11, 2015, at 14:23, Durham Goode <durham at fb.com> wrote:
> 
> 
> 
> On 12/11/15 9:46 AM, Augie Fackler wrote:
>> On Fri, Dec 11, 2015 at 12:22 PM, Durham Goode <durham at fb.com> wrote:
>>>> What's up with the sshpeer.sshpeer isinstance checks? Nothing else is
>>>> assigning to this attribute right now, so it shouldn't matter...
>>> The http peer is also stored in self.remoteserver, so the http tests fails
>>> without that check.
>> 
>> Ah. That's a little groty. I think where I'd go with this is look for
>> self.remoteserver.cleanup() and call that if it exists in the
>> teardown.
>> 
>> I'm less sure what to do about the bit where we ensure the connection
>> is alive - perhaps we should expand the peer API to include something
>> useful on that front instead of just having random isinstance checks
>> sprinkled in various bits of code. That's more of an idea for later, I
>> guess, but it seems like something we should clean up before too long.
> K, I made the requested changes (moved to _connect() function, keyed cleanup off of 'cleanup' attribute), and pushed it.
> 
> I'm going to be doing a bit of work in remotefilelog in the near future, so I'll look into refactoring some of this stuff (like adding a isalive() on ssh/httppeer's)

Awesome. I'm happy to chat about things and/or do reviews if it's helpful.

AF