Re: Timeout::Error

Nobuyoshi Nakada wrote:
> Hi,
> 
> At Wed, 6 Feb 2008 18:05:45 +0900,
> Jeremy Thurgood wrote in [ruby-core:15374]:
>> What was the change? I currently lack the time and resources to go 
>> digging through commit logs for it.
> 
> TimeoutError was a subclass of StandardError till 1.6, but now
> it inherits Interrupt which is not a subclass of StandardError.

Any idea why this changed?

>> Thanks for the patch. It only fixes the symptom, however, not the 
>> problem. I still have to explicitly handle timeouts in (for example) 
>> HTTP requests instead of simply noting a general failure (in a bare 
>> "rescue" clause), logging the exception and retrying a couple of times 
>> before giving up and moving on to something else. For most of my use 
>> cases, this is exactly how I would handle HTTP requests in long-running 
>> services that require me to keep state.
> 
> I'm just curious, do you run long-running services in irb?

No, I first noticed this in irb. I write long-running services in Ruby 
that may unknowingly invoke timeout() deep within libraries and thus 
crash. Try the following piece of code, imagining that the timeout block 
actually does something with a possibly slow network:

#!/usr/bin/env ruby

require 'timeout'

while true
   begin
     randnum = (rand*10).to_i
     Timeout::timeout(6) do
       puts "Doing something for #{randnum} seconds..."
       sleep(randnum)
     end
     puts "I did something without timing out!"
   rescue
     puts "Oops, something went wrong. I'll try again."
   end
end


--J

--