<div dir="ltr">..... why are you using exceptions are part of normal control flow for actions that aren't some manner of thread timeout or otherwise exceptional?<div><br></div><div>on a more important note... its important to note that youre focusing on the *profiling build* notion of call stack, rather than the (still moderately in progress) dwarf stack trace work thats still on going. The stack trace spamming issue you're alluding to that arises in the +RTS -xc -RTS profiling stack traces should not ever happen with dwarf stack traces.</div><div><br></div><div>what precludes changing all exceptions in base/userland to having proper stack traces associated with them as an alternative design that addresses that same issues. I ask this because any changes to base are only going to be happening in future GHCs, and thus any discussions about changing exceptions in base really need to be forward looking with respect to parallel materially ongoing work in GHC</div><div><br></div><div>for those who wanna read up on whats currently afoot in stack trace land for GHC, let me share the following links</div><div><br></div><div><a href="https://ghc.haskell.org/trac/ghc/ticket/3693#comment:75">https://ghc.haskell.org/trac/ghc/ticket/3693#comment:75</a><br></div><div><a href="https://ghc.haskell.org/trac/ghc/wiki/DWARF">https://ghc.haskell.org/trac/ghc/wiki/DWARF</a><br></div><div><a href="http://arashrouhani.com/papers/master-thesis.pdf">http://arashrouhani.com/papers/master-thesis.pdf</a><br></div><div><br></div></div><div class="gmail_extra"><br><div class="gmail_quote">On Tue, Apr 21, 2015 at 11:43 PM, Michael Sloan <span dir="ltr"><<a href="mailto:mgsloan@gmail.com" target="_blank">mgsloan@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">Davean's proposal is essentially the same as mine, restarted and specialized to callstacks. So, I'm not sure why it would make you vote against this.<br><br>This does more than just stack traces, and is independent of the source of call stacks. What kind of mechanism are you envisioning that would make it redundant to include the callstack with the exception?<br><br>I can imagine setting a thread local variable to the "last callstack of raise#". However, this does not allow us to accumulate callstacks when the exception is rethrown. Often you care more about the initial throw of the exception, rather than the most recent one.<br><br>For example, with `+GHC -xc` output, when an exception is caught and rethrown, you'll see the callstacks of all the places it's thrown. This output is rather terrible for real world debugging, though, because it doesn't actually tell you what the exception is. When dealing with a concurrent system where some exceptions occur as part of normal operation, this becomes nightmareish. This proposal solves that problem. If you see even the rough possibility of another solution to this problem, please do tell.<span class="HOEnZb"><font color="#888888"><div><br></div><div>-Michael</div></font></span></div><div class="HOEnZb"><div class="h5"><div class="gmail_extra"><br><div class="gmail_quote">On Tue, Apr 21, 2015 at 7:26 PM, Carter Schonwald <span dir="ltr"><<a href="mailto:carter.schonwald@gmail.com" target="_blank">carter.schonwald@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">in the context of davean's proposal (which i'm still digesting), i'm gonna go -1 on this one.<div><br></div><div>i'm really leery of commiting to any changes to our exception machinery until the dwarf stack trace tooling and associated RTS/exception interaction support is a bit more mature, because i think a lot of other approaches / changes to ghc / base have been driven by the lack of cheap stack traces. This proposal crosses that line, at least for me ;) </div></div><div><div><div class="gmail_extra"><br><div class="gmail_quote">On Tue, Apr 21, 2015 at 6:24 PM, Michael Sloan <span dir="ltr"><<a href="mailto:mgsloan@gmail.com" target="_blank">mgsloan@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">No, this proposal is not specifically about stack traces, that is just one of the usecases. Instead, this is about a general mechanism for including extra information with exceptions. The core of this proposal is still relevant even if the behavior of error / throw / throwTo / etc remain unchanged.<div><br><div><div>I'm not familiar with how the new dwarf stuff will interact with throwing / displaying exceptions. It seems like this would require having the debugger break at the throw site, and exceptions would still lack stack traces. Having informative stack traces is quite orthogonal to having a good place to store them.<br><br>Note that in my original proposal text I mentioned that this is agnostic of the particular source of the stack trace. In particular, this could be used with a profiling stack trace, implicit callstack, or, indeed, these traces via dwarf.</div><span><font color="#888888"><div><br></div><div>-Michael </div></font></span></div></div></div><div><div><div class="gmail_extra"><br><div class="gmail_quote">On Tue, Apr 21, 2015 at 3:06 PM, Carter Schonwald <span dir="ltr"><<a href="mailto:carter.schonwald@gmail.com" target="_blank">carter.schonwald@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">On a more important note: assuming ghc 7.12 has support for informative stack traces via dwarf by default, wouldn't that eliminate the need for this proposal? Namely : there perhaps should be some reasonable way to talk about concatting stack traces perhaps?<div><br></div><div>Phrased differently: how is the info that should perhaps be in informative stack traces not subsuming the info of this proposal?<div><div><span></span><br><br>On Tuesday, April 21, 2015, Michael Sloan <<a href="mailto:mgsloan@gmail.com" target="_blank">mgsloan@gmail.com</a>> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr"><div>Ah, but it looks like Niklas does have a patch which adds implicit locations to such functions: <a href="https://phabricator.haskell.org/D861" style="font-size:12.8000001907349px" target="_blank">https://phabricator.haskell.org/D861</a><br><br>However, there are some issues with changing the API of these functions: <a href="https://phabricator.haskell.org/D861#23250" target="_blank">https://phabricator.haskell.org/D861#23250</a></div><br>(as mentioned in the "Backporting srcLoc to the GHC 7.10 branch" thread)<div class="gmail_extra"><br><div class="gmail_quote">On Tue, Apr 21, 2015 at 2:04 PM, Michael Sloan <span dir="ltr"><<a>mgsloan@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex"><div dir="ltr">Hmm, that patch doesn't appear to add stack traces to 'Prelude.error', which is what Carter wants here. Also, I think it would be done with profiling callstacks rather than implicit callstacks. But it's certainly also useful to have functions which do the same with implicit callstacks!<div><div><br><div class="gmail_extra"><br><div class="gmail_quote">On Tue, Apr 21, 2015 at 1:55 PM, Evan Laforge <span dir="ltr"><<a>qdunkan@gmail.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">Maybe I'm missing something, but isn't this already implemented?<br>
<br>
<a href="https://phabricator.haskell.org/D578" target="_blank">https://phabricator.haskell.org/D578</a><br>
<div><div><br>
On Tue, Apr 21, 2015 at 1:37 PM, Michael Sloan <<a>mgsloan@gmail.com</a>> wrote:<br>
> On Thu, Apr 16, 2015 at 8:08 PM, Carter Schonwald<br>
> <<a>carter.schonwald@gmail.com</a>> wrote:<br>
>><br>
>> if you can patch prelude error to include stack traces, i will owe you a<br>
>> >=1 beer each at the next two icfps. Thats all i want for christmas. :)<br>
><br>
><br>
> Sounds good! No promises, but I'll be giving this a try soon. Looking<br>
> forward to ICFP beers either way :D<br>
><br>
>> i can't speak for how a different patch might work out, because thats not<br>
>> what I'd tried at the time. If you have a go, please share the results!<br>
>> -Carter<br>
>><br>
>> On Wed, Apr 15, 2015 at 12:22 AM, Michael Sloan <<a>mgsloan@gmail.com</a>> wrote:<br>
>>><br>
>>> Hi Carter!<br>
>>><br>
>>> Interesting! This thread, right?<br>
>>> <a href="https://mail.haskell.org/pipermail/libraries/2014-December/024429.html" target="_blank">https://mail.haskell.org/pipermail/libraries/2014-December/024429.html</a><br>
>>><br>
>>> I haven't tried this as a patch to base, but I'm certain that the core of<br>
>>> the proposal has no extra dependencies. Note that the proposal isn't about<br>
>>> stack traces in particular - that's just one application of being able to<br>
>>> throw exceptions with extra information.<br>
>>><br>
>>> Even if `throwTo` isn't modified to throw exceptions with stack traces,<br>
>>> this functionality could be provided outside of `Control.Exception` (though,<br>
>>> that does seem like the right place to put it). I'm surprised that the<br>
>>> circularity was so problematic, though. Why isn't it sufficient to have an<br>
>>> hs-boot file for `GHC.Stack`, which exports `currentCallStack`?<br>
>>><br>
>>> -Michael<br>
>>><br>
>>> On Tue, Apr 14, 2015 at 7:55 PM, Carter Schonwald<br>
>>> <<a>carter.schonwald@gmail.com</a>> wrote:<br>
>>>><br>
>>>> Hey Michael,<br>
>>>> I actually proposed something along these lines that got OK'd by<br>
>>>> libraries early this past fall, the main challenge we hit was actually doing<br>
>>>> the enginering to add the stack traces to exceptions! theres some nasty<br>
>>>> module cycles in base that happen when you try to weave things around so<br>
>>>> that the standard error "message here" call includes some stack trace info.<br>
>>>> Have you tried to do that simple starter patch to base?<br>
>>>><br>
>>>> Chris Allen and I spent like 2 days trying to get it to work and just<br>
>>>> gave up because of the cycles. We (and others) would probably love some<br>
>>>> headway on that front.<br>
>>>><br>
>>>> Theres also some in progress work to use the dwarf debugging info data<br>
>>>> in >7.10 to provide useful stack traces in the default builds for GHC afaik,<br>
>>>> 'cause the stack trace functionality you're pointing at currenlty only work<br>
>>>> on profiled builds<br>
>>>><br>
>>>> cheers<br>
>>>> -Carter<br>
>>>><br>
>>>> On Tue, Apr 14, 2015 at 2:38 PM, Michael Sloan <<a>mgsloan@gmail.com</a>><br>
>>>> wrote:<br>
>>>>><br>
>>>>> Control.Exception currently lacks a good way to supply extra<br>
>>>>> information along with exceptions. For example, exceptions could be<br>
>>>>> thrown along with their callstack[1] or implicit stack[2], but we have<br>
>>>>> no generic way to include this information with exceptions.<br>
>>>>><br>
>>>>> Proposed Solution<br>
>>>>> =================<br>
>>>>><br>
>>>>> The proposed solution is to add a list of `SomeExceptionInfo` to the<br>
>>>>> `SomeException` datatype. This list stores additional information<br>
>>>>> about the exception. These `ExceptionInfo` instances use a mechanism<br>
>>>>> which is pretty much identical to the dynamic way the `Exception` type<br>
>>>>> works:<br>
>>>>><br>
>>>>> data SomeException = forall e . Exception e =><br>
>>>>> SomeExceptionWithInfo e [SomeExceptionInfo]<br>
>>>>><br>
>>>>> data SomeExceptionInfo = forall a . ExceptionInfo a =><br>
>>>>> SomeExceptionInfo a<br>
>>>>><br>
>>>>> class Typeable a => ExceptionInfo a where<br>
>>>>> displayExceptionInfo :: a -> String<br>
>>>>><br>
>>>>> addExceptionInfo<br>
>>>>> :: (ExceptionInfo a, Exception e)<br>
>>>>> => a -> e -> SomeException<br>
>>>>> addExceptionInfo x (toException -> SomeExceptionWithInfo e xs) =<br>
>>>>> SomeExceptionWithInfo e (SomeExceptionInfo x : xs)<br>
>>>>><br>
>>>>> `ExceptionInfo` lacks the to / from functions that `Exception` has,<br>
>>>>> because I don't see much point in supporting a hierarchy for exception<br>
>>>>> info. The `Typeable` superclass constraint supplies the necessary<br>
>>>>> casting.<br>
>>>>><br>
>>>>> `SomeExceptionInfo` could validly instead just use the constraint<br>
>>>>> `(Typeable a, Show a)`. However, I believe it's good to have a new<br>
>>>>> class for this so that:<br>
>>>>><br>
>>>>> * The user can specify a custom `displayExceptionInfo`<br>
>>>>> implementation, for when this extra info is presented to the user.<br>
>>>>> This function would be invoked by the `show` implementation for<br>
>>>>> `SomeException`.<br>
>>>>><br>
>>>>> * Types need to opt-in to be usable with `SomeExceptionInfo`.<br>
>>>>> Similarly to exceptions, I imagine that a type with a<br>
>>>>> `ExceptionInfo` instance won't be used for anything but acting as<br>
>>>>> such an annotation. Having a class for this allows you to ask GHCI<br>
>>>>> about all in-scope exception info types via `:info ExceptionInfo`.<br>
>>>>><br>
>>>>> Backwards Compatibility<br>
>>>>> =======================<br>
>>>>><br>
>>>>> GHC 7.10 adds support for bidirectional pattern synonyms. This means<br>
>>>>> that this change could be made without breaking code:<br>
>>>>><br>
>>>>> pattern SomeException x <- SomeExceptionWithInfo x _ where<br>
>>>>> SomeException x = SomeExceptionWithInfo x []<br>
>>>>><br>
>>>>> Note that consumers of this do not need to enable `-XPatternSynonyms`.<br>
>>>>><br>
>>>>> Applications<br>
>>>>> ============<br>
>>>>><br>
>>>>> Callstacks<br>
>>>>> ----------<br>
>>>>><br>
>>>>> As mentioned at the beginning, this can be used to add callstacks to<br>
>>>>> exceptions:<br>
>>>>><br>
>>>>> newtype ExceptionCallStack =<br>
>>>>> ExceptionCallStack { unExceptionCallStack :: [String] }<br>
>>>>> deriving Typeable<br>
>>>>><br>
>>>>> instance ExceptionInfo ExceptionCallStack where<br>
>>>>> displayExceptionInfo = unlines . unExceptionCallStack<br>
>>>>><br>
>>>>> throwIOWithStack :: Exception e => e -> IO a<br>
>>>>> throwIOWithStack e = do<br>
>>>>> stack <- currentCallStack<br>
>>>>> if null stack<br>
>>>>> then throwIO e<br>
>>>>> else throwIO (addExceptionInfo (ExceptionCallStack stack)<br>
>>>>> e)<br>
>>>>><br>
>>>>> I see little downside for making something like this the default<br>
>>>>> implementation `throwIO`. Each rethrowing of the `SomeException`<br>
>>>>> would add an additional stacktrace to its annotation, much like the<br>
>>>>> output of `+RTS -xc`. Unlike this debug output, though, the<br>
>>>>> stacktraces would be associated with the exception, rather than just<br>
>>>>> listing locations that exceptions were thrown. This makes it<br>
>>>>> tractable to debug exceptions that occur in concurrent programs, or in<br>
>>>>> programs which frequently throw exceptions during normal functioning.<br>
>>>>><br>
>>>>> Throwing Exceptions in Handlers<br>
>>>>> -------------------------------<br>
>>>>><br>
>>>>> Example:<br>
>>>>><br>
>>>>> main =<br>
>>>>> throwIO InformativeErrorMessage `finally`<br>
>>>>> throwIO ObscureCleanupIssue<br>
>>>>><br>
>>>>> While `InformativeErrorMessage` got thrown, the user doesn't see it,<br>
>>>>> since `ObscureCleanupIssue` is thrown instead. This causes a few<br>
>>>>> issues:<br>
>>>>><br>
>>>>> 1. If the exception is handled by the default handler and yielded to<br>
>>>>> the user, then the more informative error is lost.<br>
>>>>><br>
>>>>> 2. Callers who expect to catch the "Informative error message" won't<br>
>>>>> run their handlers for this exception type.<br>
>>>>><br>
>>>>> Problem 1 can now easily be resolved by adding some info to the<br>
>>>>> exception:<br>
>>>>><br>
>>>>> data ExceptionCause = ExceptionCause<br>
>>>>> { unExceptionCause :: SomeException }<br>
>>>>> deriving Typeable<br>
>>>>><br>
>>>>> instance ExceptionInfo ExceptionCause where<br>
>>>>> displayExceptionInfo fe =<br>
>>>>> "thrown while handling " ++<br>
>>>>> displayException (unExceptionCause fe)<br>
>>>>><br>
>>>>> catch :: Exception e => IO a -> (e -> IO a) -> IO a<br>
>>>>> catch f g = f `oldCatch` handler<br>
>>>>> where<br>
>>>>> handler ex = g ex `oldCatch` \(ex' :: SomeException) -><br>
>>>>> throwIO (addExceptionInfo info ex')<br>
>>>>> where<br>
>>>>> info = ExceptionCause (toException ex)<br>
>>>>><br>
>>>>> This implementation of `catch` is written in a backwards-compatible<br>
>>>>> way, such that the exception thrown during finalization is still the<br>
>>>>> one that gets rethrown. The "original" exception is recorded in the<br>
>>>>> added info. This is the same approach used by Python 3's<br>
>>>>> `__context__` attribute[3]. This was brought to my attention in a<br>
>>>>> post by Mike Meyer[4], in a thread about having bracket not suppress<br>
>>>>> the original exception[5].<br>
>>>>><br>
>>>>> This doesn't directly resolve issue #2, due to this backwards<br>
>>>>> compatibility. With the earlier example, a `catch` handler for<br>
>>>>> `InformativeErrorMessage` won't be invoked, because it isn't the<br>
>>>>> exception being rethrown. This can be resolved by having a variant of<br>
>>>>> catch which instead throws the original exception. This might be a<br>
>>>>> good default for finalization handlers like `bracket` and `finally`.<br>
>>>>><br>
>>>>> Asynchronous Exceptions<br>
>>>>> -----------------------<br>
>>>>><br>
>>>>> Currently, the only reliable way to catch exceptions, ignoring async<br>
>>>>> exceptions, is to fork a new thread. This is the approach used by the<br>
>>>>> enclosed-exceptions[6] package. I think it's quite ugly that we need<br>
>>>>> to go to such lengths due to the lack of one bit of information about<br>
>>>>> the exception! This would resolve ghc trac #5902[7].<br>
>>>>><br>
>>>>> base-4.7 added the `SomeAsyncException` type, but this doesn't enforce<br>
>>>>> anything. Any exception can be thrown as a sync or async exception.<br>
>>>>> Instead, we ought to have a reliable way to know if an exception is<br>
>>>>> synchronous or asynchronous. Here's what this would look like:<br>
>>>>><br>
>>>>> data IsAsync = IsAsync<br>
>>>>> deriving (Typeable, Show)<br>
>>>>><br>
>>>>> instance ExceptionInfo IsAsync where<br>
>>>>> displayExceptionInfo IsAsync = "thrown asynchronously"<br>
>>>>><br>
>>>>> throwTo :: Exception e => ThreadId -> e -> IO ()<br>
>>>>> throwTo tid = oldThrowTo tid . addExceptionInfo IsAsync<br>
>>>>><br>
>>>>> The details of this get a bit tricky: What happens if `throwIO` is<br>
>>>>> used to rethrow a `SomeException` which has this `IsAsync` flag set?<br>
>>>>> I'm going to leave out my thoughts on this for now as the interactions<br>
>>>>> between unsafePerformIO and the concept of "rethrowing" async<br>
>>>>> exceptions. Such details are explained in a post by Edsko de Vries[8]<br>
>>>>> and ghc trac #2558[9].<br>
>>>>><br>
>>>>> Issue: fromException loses info<br>
>>>>> ===============================<br>
>>>>><br>
>>>>> I can think of one main non-ideal aspect of this proposal:<br>
>>>>><br>
>>>>> Currently, the `toException` and `fromException` methods usually form<br>
>>>>> a prism. In other words, when `fromException` yields a `Just`, you<br>
>>>>> should get the same `SomeException` when using `toException` on that<br>
>>>>> value.<br>
>>>>><br>
>>>>> For example,<br>
>>>>><br>
>>>>> fail "testing 1 2 3" `catch` \(ex :: SomeException) -> throwIO ex<br>
>>>>><br>
>>>>> is equivalent to<br>
>>>>><br>
>>>>> fail "testing 3 4 5" `catch` \(ex :: IOError) -> throwIO ex<br>
>>>>><br>
>>>>> However, with exception info added to just `SomeException`, and no<br>
>>>>> changes to existing `Exception` instances, this<br>
>>>>> doesn't hold. Exceptions caught as a specific exception type get<br>
>>>>> rethrown with less information.<br>
>>>>><br>
>>>>> One resolution to this is be to add `[SomeExceptionInfo]` as a field<br>
>>>>> to existing `Exception` instances. This would require the use of<br>
>>>>> non-default implementations of the `toException` and `fromException`<br>
>>>>> instances.<br>
>>>>><br>
>>>>> Another approach is to have variants of `catch` and `throw` which also<br>
>>>>> pass around the `[SomeExceptionInfo]`.<br>
>>>>><br>
>>>>> [1]<br>
>>>>> <a href="https://hackage.haskell.org/package/base-4.8.0.0/docs/GHC-Stack.html#currentCallStack" target="_blank">https://hackage.haskell.org/package/base-4.8.0.0/docs/GHC-Stack.html#currentCallStack</a><br>
>>>>> [2]<br>
>>>>> <a href="https://ghc.haskell.org/trac/ghc/wiki/ExplicitCallStack/ImplicitLocations" target="_blank">https://ghc.haskell.org/trac/ghc/wiki/ExplicitCallStack/ImplicitLocations</a><br>
>>>>> [3] <a href="https://www.python.org/dev/peps/pep-3134/" target="_blank">https://www.python.org/dev/peps/pep-3134/</a><br>
>>>>> [4]<br>
>>>>> <a href="https://mail.haskell.org/pipermail/haskell-cafe/2014-July/114987.html" target="_blank">https://mail.haskell.org/pipermail/haskell-cafe/2014-July/114987.html</a><br>
>>>>> [5]<br>
>>>>> <a href="https://mail.haskell.org/pipermail/haskell-cafe/2014-July/114986.html" target="_blank">https://mail.haskell.org/pipermail/haskell-cafe/2014-July/114986.html</a><br>
>>>>> [6] <a href="https://hackage.haskell.org/package/enclosed-exceptions" target="_blank">https://hackage.haskell.org/package/enclosed-exceptions</a><br>
>>>>> [7] <a href="https://ghc.haskell.org/trac/ghc/ticket/5902" target="_blank">https://ghc.haskell.org/trac/ghc/ticket/5902</a><br>
>>>>> [8] <a href="http://www.edsko.net/2013/06/11/throwTo/" target="_blank">http://www.edsko.net/2013/06/11/throwTo/</a><br>
>>>>> [9] <a href="https://ghc.haskell.org/trac/ghc/ticket/2558" target="_blank">https://ghc.haskell.org/trac/ghc/ticket/2558</a><br>
>>>>><br>
>>>>> _______________________________________________<br>
>>>>> Libraries mailing list<br>
>>>>> <a>Libraries@haskell.org</a><br>
>>>>> <a href="http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries" target="_blank">http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries</a><br>
>>>>><br>
>>>><br>
>>><br>
>><br>
><br>
><br>
> _______________________________________________<br>
> Libraries mailing list<br>
> <a>Libraries@haskell.org</a><br>
> <a href="http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries" target="_blank">http://mail.haskell.org/cgi-bin/mailman/listinfo/libraries</a><br>
><br>
</div></div></blockquote></div><br></div></div></div></div>
</blockquote></div><br></div></div>
</blockquote></div></div></div>
</blockquote></div><br></div>
</div></div></blockquote></div><br></div>
</div></div></blockquote></div><br></div>
</div></div></blockquote></div><br></div>