[racket-dev] Problem with latest nightly build

Thanks for the report!

It's an optimizer bug. The bytecode compiler is currently transforming

   (let ([h (box 0.0)])
     (lambda () h))

to

  (lambda () (box 0.0))

which is obviously wrong. I'll work on that.


At Mon, 15 Aug 2011 08:48:41 -0600, Doug Williams wrote:
> I downloaded the latest nightly build to test a fix that Matthew pushed.
> That fix (a refresh problem) worked, but when I tried to run some
> numerically intensive things - like the test for the science collection - I
> ran into problems. The science test hangs on the last differential equation
> test - eventually it runs out of memory. Some other examples seg fault in
> the garbage collector. I tried both 32 and 64 bit windows versions and they
> both had the same symptoms. These all work fine in 5.1.2. I tried a nightly
> build from yesterday (5.1.2.4) and one from this morning (5.1.3.1) and they
> both had the problem.
> 
> Sorry I can't be more specific. You can require the science collection from
> PLaneT - williams/science/science - and run science.test.ss. That one fails
> every time for me by running out of memory eventually - I typically have
> memory limited to 4096MB on a 64-bit build or 2048 on a 32-bit build. I also
> tried some of the simulation examples that use the same differential
> equation solver - they're in williams/simulation/simulation. For example,
> model-2a.ss will sometime run fine and sometimes fail with a seg fault.
> Another one that fails sometimes and runs others is interactions.ss in the
> simulation collection.
> 
> Let me know if I can get you any other information on the problem.
> 
> Doug
> _________________________________________________
>   For list-related administrative tasks:
>   http://lists.racket-lang.org/listinfo/dev