From: Matthew Bloch <mattbee@...>
Date: 2002-07-01T02:08:45+09:00
Subject: Re: Getting a full backtrace from Thread#join?

Matthew Bloch wrote:

> Here's a program that demonstrates some unhelpful output:
> 
>   def countdown(n)
>     if n == 0
>       raise Exception.new
>     else
>       countdown(n-1)
>     end
>   end
> 
>   a = Thread.new { countdown(5) }
> 
>   begin
>     a.join
>   rescue Exception => ex
>     print "#{ex}\n"
>     print ex.backtrace.join("\n")
>     print "\n"
>   end
> 
> The backtrace I see looks like this:
> 
>   Exception
>   bt:3:in `countdown'
>   bt:12:in `join'
>   bt:12
> 
> Problem is that only the first line of the backtrace belongs to the thread
> which threw the exception (whereas I should see 5 nested calls to
> 'countdown').  I'm trying to weed out an exception in a multithreaded
> application and one line of backtrace isn't enough; is there any way of
> getting hold of the thread's full backtrace?

Has anyone got a better solution than overriding Thread and using this kind 
of setup?

class EThread < Thread
  def initialize(*args)
    super(*args) do |a|
      begin
        yield(*a)
      rescue Exception => e
        # deal with exception here, not outside
        #
        Log.error "thread #{self} exception '#{e.type} : #{e.message}'"     
        Log.backtrace(e)
        raise
      end
    end
  end
end

-- 
Matthew