timeouts

I’m using octave_kernel, but I’m now certain there is no (easy) way to fix my problem without changing process_metakernel in Calysto. When I run a command that takes a long time to return a prompt, or any kind of text for that matter, I get the dreaded pexpect 30-second default timeout.

A “fix” for this was committed almost a year ago (8b4429ffafbc201b4e8af5bcbe00ac647a8176f0). However, the critical part (timeout=None) was reverted barely a week later (timeout=-1, d9c975538b30061a7a6a62bedc4e8b2fc4447f88), along with changes to pexpect itself that changed the behavior of timeout=None to just use the default timeout of 30 seconds.

I have to admit I understand very little of this, but I suspect the author(s) of these changes consider the original bug to be only an issue when there is a continuous stream of output from a long-running process. I would ask that they consider the case when a process runs quietly for a long time. I realize this is probably an edge case, but at least for the kind of work I do with Octave, it is not all that uncommon.

Maybe there is a “fix” that at least allows the child kernel (octave_kernel in this case) to specify the timeout. As it stands, my only options are to manually modify the pexpect files, or to modify my Octave code to spit out a stream of output that prevents the timeout from triggering. Neither of these are very satisfying.

[blink1073]

I’ve been meaning to update the semantics of the %plot magic similarly, I will try and do both this weekend.

[blink1073]

I was in the process of adding this, and realized we should not have a timeout at all. I am going to look into using pexpect 4.2 as a dependency, which has Windows support.