If the PROTOCOLINFO response has an unrecognized method of authentication (such as 'SAFECOOKIE') we raise an IOError which, in turn, breaks connect(). Here's the offending bit of code...
CCing Roger and Linus, too, because they run Torperf instances and should probably be informed that PyTorCtl will be incompatible with Tor 0.2.2.36 when that is released.
Ah ha. Hm. Is there any way to fix this point about 0.2.2.36? New stable minor releases shouldn't break, say, all the arm users if we can help it. It's not too late to change our minds. :)
New stable minor releases shouldn't break, say, all the arm users if we can help it.
I should make another arm release at some point soonish to get out several other bug fixes too so I wouldn't object to bumping up those plans. Arm is already fixed and unaffected by torctl for this (as mentioned on irc I needed to duplicate this torctl functionality in order to add other unmerged torctl features like control socket support).
CCing Roger and Linus, too, because they run Torperf instances and should probably be informed that PyTorCtl will be incompatible with Tor 0.2.2.36 when that is released.
Actually, Linus and I run Torperf instances. They all use Tor 0.2.2.35. What's the suggested workaround? Not upgrade to 0.2.2.36 ever?
What's the suggested workaround? Not upgrade to 0.2.2.36 ever?
Or comment out the IOError I cited earlier. This is a one line fix, just replace...
raise IOError("Unrecognized authentication type: %s" % authInfo)
with...
pass # unrecognized auth method
Will there be a pytorctl fix? I don't like the idea of making local changes to pytorctl in the three places that are running Torperf. I also don't feel comfortable enough with the pytorctl code to write and test this patch myself.
Dude, there are so many balls in my court I am just going to order a new floor and build over them. I'm already wading through them as it is. Or... Profit? Anybody wanna buy a ball pit?
However, because the fix seems a simple hack of changing the exception to a log message, I just went ahead and pushed it. Hope it works.
Please continue to hassle aagbsn about torflow issues. If he isn't responsive for simple obvious fixes (big rewrites to address minor control port protocol changes don't count), we'll just have to find a new torflow maintainer.