On Tue, Dec 11, 2012 at 2:44 AM, Sven Van Caekenberghe <sven@stfx.eu> wrote:
Hi Eliot,
Thanks for the reply.
On 11 Dec 2012, at 02:33, Eliot Miranda <eliot.miranda@gmail.com> wrote:
> 2012/12/9 Sven Van Caekenberghe <sven@stfx.eu>
> Hi,
>
> Would it be possible to build some kind of protection against stack overflow ?
>
> note that there already is some mechanism. �The low space mechanism is supposed to protect against stack overflow but with today's memories and the difficulty of testing it, this mechanism often a) takes way too long to kick in, and b) either leaves the system in a very sluggish state or simply fails to kick in and the system crashes with an out-of-memory error.
>
> Part of the problem is in designing a mechanism which is in keeping with the reflective nature of the system. �What I mean is that we now have two different VMs, the Interpreter and the STack/Cog VMs. �First one wants a check which is cheap. �That means a check which isn't run on every send. �In the interpreter a natural time to check for recursion would be on e.g. fullGC, checking for repetition in the current process's context chain. �The the Stack/Cog VMs the natural time is on evacuating a stack page when the stack zone is full. �But these are two different mechanisms, both of which are deep in the VM, neither of which can be conveniently intercepted from Smalltalk, unlike e.g. doesNotUnderstand:.
>
> Personally I like the idea of reviving the low space mechanism and maintaining it properly. �For example, if the low space mechanism causes the low space semaphore to fire often, because the limit was set only a little way above the current memory size, and not, as it is now, a little below the absolute limit, then when it fires, the low space process could inspect the current process's stack (e.g. via accelerated via a primitive for performance) and see how deep it is, and possibly whether there is recursion (e.g. a bag of context methods would soon reveal high-frequency methods, and then those contexts could be inspected for repetition, but KISS says just have a per-process limit on the maximum depth of stack).
Yes, getting a low space warning sooner and then deal with that intelligently at the image level sounds like a good solution. It is probably more portable, requires miminal VM collaboration, allows for many behaviours as a reaction and would catch other problems than just stack overflow.
Is this already possible today ?
Wow, I'm amazed. �The low space limit in the VM has been ripped out �So no, it is not possible, but it would be easy to put back. �There is a variable in ObjectMemory (the VM's heap manager/garbage collector class) that defines the low-space limit, but there is no setter. �Back in the day (and still in VisualWorks) there was a primitiveSystemDictionary methods for memory spacesignal: aSemaphore atOopsLeft: numOops wordsLeft: numWords�"Tell the object memory to signal the Semaphore when either the number�of object pointers remaining drops below numOops, or the number of �words in the object space remaining drops below numWords. �Fail if the �frist argument is neither a Semaphore nor nil. �Fail if numOops is not a�16-bit Integer, or if numWords is not a 32-bit LargePositiveInteger. �Essential. �See Object documentation whatIsAPrimitive."<primitive: 116>self primitiveFailedBut this and its support is gone. �Hmm. �Hmmph :)�Preferrably, dynamic control over the limit would be nice: in a headless server you want other behavior than during interactive development.Right. �The low space process would readjust the value every time low space fired. �But one could (can) call the primitive at any time to install a new limit.
Sven
> This is actually the most common situation for loop/cycle that either results in an unusable image or a crash.
> Interrupting with CMD-. is not always possible.
> A possible solution could be to just limit the stack size and throw an exception.
>
> For example, in LispWorks, it goes like this:
>
> CL-USER 5 > (defun foo () (cons (random 100) (foo)))
> FOO
>
> CL-USER 6 > (foo)
>
> Stack overflow (stack size 48128).
> � 1 (continue) Extend stack by 50%.
> � 2 (abort) Return to level 0.
> � 3 Return to top loop level 0.
>
> Type :b for backtrace or :c <option number> to proceed.
> Type :bug-form "<subject>" for a bug report template or :? for other options.
>
> CL-USER 7 : 1 > :bq
>
> ERROR <- RUNTIME:BAD-ARGS-OR-STACK <- LENGTH <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO <- FOO
> <- FOO <- FOO <- EVAL <- CAPI::CAPI-TOP-LEVEL-FUNCTION <- CAPI::INTERACTIVE-PANE-TOP-LOOP <- MP::PROCESS-SG-FUNCTION
>
> Obviously, this is possible, the question is, can it be done easily and without paying an unbearable performance price ?
>
> Sven
>
> --
> Sven Van Caekenberghe
> http://stfx.eu
> Smalltalk is the Red Pill
>
>
>
>
>
>
>
> --
> best,
> Eliot
>
--
Sven Van Caekenberghe
http://stfx.eu
Smalltalk is the Red Pill
--
best,Eliot