[Date Index] [Thread Index] [Author Index]

Re: Re: Compatibility woes

• To: mathgroup at smc.vnet.net
• Subject: [mg76202] Re: [mg76139] Re: Compatibility woes
• From: Andrzej Kozlowski <akoz at mimuw.edu.pl>
• Date: Thu, 17 May 2007 06:15:13 -0400 (EDT)
• References: <f2btsp$iha$1@smc.vnet.net> <200705160951.FAA00779@smc.vnet.net>

If you check the documentation for Compile in 6.0 you will see at the  
bottom o fht epage:

New in 2  |  Last modified in 3

I, for one, find this disappointing. When I read your message I was  
briefly excited, thinking: Great! At last Compile got modified and  
that must mean that more things can be compiled, perhaps even  
patterns or at least non-tensor arrays. Alas, there has been no  
change and your problems must be caused by something else. :-((

Andrzej Kozlowski

On 16 May 2007, at 18:51, Mark Adler wrote:

> Thanks to Bhuvanesh, I got the package to load (by being explicit
> about Developer` for those functions, instead of declaring it as
> Needed in the BeginPackage[]).  So then I discover that Compile[] no
> longer works in 6.0 the same as it did in 5.2.  I replaced it with
> Function[], modifying the arguments appropriately, and the package
> works, albeit slower than before.
>
> So that's yet another thing that worked just peachy in 5.2, but no
> longer works in 6.0.  I doubt that I have plumbed the depths of the
> incompatibilities, since the rate of discovery is not decreasing.
>
> Mark Adler
>
>
> For the curious, the errors look like:
>
> CompiledFunction::cfse: Compiled expression
>     MarsKSAM`Private`base[MarsKSAM`Private`k][-1.05] should be a
>     machine-size real number. More...
>
> CompiledFunction::cfex: External evaluation error at instruction 87;
>     proceeding with uncompiled evaluation. More...
>
> This all worked perfectly in 5.2.
>
>

• References:
  • Re: Compatibility woes
    • From: Mark Adler <madler@alumni.caltech.edu>