[Openmcl-devel] ABCL and early-unwinding in .net

Hamilton Link helink at sandia.gov
Thu Jun 12 14:19:34 PDT 2008


Brian Mastenbrook wrote:
> On Jun 12, 2008, at 1:16 PM, Andrew Shalit wrote:
>
>   
>> As for the restrictions on developers: It's true those would seem to
>> disallow Lisp, smalltalk, a JVM, etc.  This may be a firm long-term
>> policy, or it might just be an interim artifact of Apple figuring out
>> their security measures.  All iPhone applications need to be signed,
>> and signatures are generated by Apple for approved developers.  I'm
>> not sure how that could work in the face of interpreters or compilers
>> residing on the iPhone.  But maybe they'll figure out a way to let it
>> happen for select tools.
>>     
>
> Since Apple's sandboxing implementation seems to be laughable at best,
> they no doubt want to ensure that any interpreter does its own
> sandboxing well enough to be trusted. I'd imagine Adobe and Sun could
> get an exemption for Flash and Java. Squeak or any of the existing CL
> implementations (aside from ABCL)... probably not so easy. A native
> Lisp environment may also not be an easy sell as the native compilers
> tend to break on major OS revisions.
>
> The most interesting possibility to me would be Microsoft putting
> Silverlight and the CLR/DLR on the iPhone. An environment which hosts
> Python and Ruby could also no doubt host an acceptable implementation
> of Common Lisp. I chatted with Gary Byers briefly about this at
> ILC'05. Many of the issues that the Common Larceny developers had
> getting Scheme on the CLR (efficient call/cc and tail-calling) are not
> an issue for CL, and the early-unwind exceptions problem has already
> been addressed in ABCL.

Hold the phone, how did ABCL circumvent the .net exception unwinding 
process?  That's great news, everything I could find indicated there 
wasn't much of a way to do it short of using the .net runtime's debugger 
hooks in a separate bit of code that talked down into the .net-hosted 
lisp (and that seemed like a huge pain).  Do you know of a good link 
covering the details?

h







More information about the Openmcl-devel mailing list