The exception object should have a full stack trace in it - how are you printing your exceptions?
Sean On Jun 2, 2014, at 12:28 PM, Lee Spector <lspec...@hampshire.edu> wrote: > In my single-threaded code, exceptions stop execution and print a stack > backtrace. This is less informative than I would like [1], but sometimes it > provides crucial clues. > > My multi-threaded code generally uses agents (just because that's the best > way I've found to get something that behaves like pmap but with better > multicore processor utilization [2]). By default, exceptions in agents are > silent, which is a default that I've always found to be perplexing. But if I > pass an :error-handler argument to agent (as in [2] below) then I can at > least see the exception... but execution doesn't halt and I don't get a > backtrace. That means that if I'm running in a multi-core mode (as I usually > am) then the first unfortunate thing is that I have to notice the printed > exception (because execution continues past it if I don't). But the even more > unfortunate thing is that I then get no backtrace. In order to get a > backtrace I have to re-run my whole system in single-threaded mode, and then > hope that I hit the same error (since my systems almost always involve a lot > of randomness). > > I figure it would be easy to put an exit in the :error-handler that would > quit the whole program, but I don't know how I could get the stack backtrace > for the triggering exception. > > Does anyone know how to get this? > > Thanks, > > -Lee > > > [1] Oh, how I long to see the values of locals up the stack in these > backtraces. I understand there are ways to get this but that they require > particular tool chains (emacs w/ritz?); my longing is for this kind of > functionality no matter how I launch my code, or at least when I launch it > via "lein run". The lack of this feature is the #1 thing I miss in Clojure > relative to Common Lisp. > > [2] Specifically, pmap won't keep all cores busy if a thread that starts > later finishes earlier than one that starts earlier. Here's the code that I > use instead of pmap, which assumes there's a boolean-valued var called > single-thread-mode: > (defn pmapall > "Like pmap but: 1) coll should be finite, 2) the returned sequence > will not be lazy, 3) calls to f may occur in any order, to maximize > multicore processor utilization, and 4) takes only one coll so far." > [f coll] > (if single-thread-mode > (doall (map f coll)) > (let [agents (map #(agent % :error-handler (fn [agnt except] (println > except))) coll)] > (dorun (map #(send % f) agents)) > (apply await agents) > (doall (map deref agents)))))
signature.asc
Description: Message signed with OpenPGP using GPGMail
