[Chicken-users] fixnum-specific math operators patch

[Will M Farr]

Hello all,

In the spirit of my last post about flonum-specific math operators,  
attached is a patch which implements the following behavior for  
Chicken's fxXXX math operators:

In safe mode: all fxXXX operators check that their arguments are  
really fixnums and throw an error if they are not.
In unsafe mode: all fxXXX operations check no argument types.

In both cases the fxXXX operators do not check for overflow (i.e. you  
can think of them as operating mod 2^n with n dependent on the system  
word size).  Right now all fxXXX operations are unsafe (in the sense  
that you can perform them on non-fixnums and get nonsense---but  
fixnum---results, not in the sense that they crash the system).

This patch would slow down code compiled in safe mode which uses the  
fxXXX operators because of the extra two checks per operation, but  
would detect stupid errors such as:

#;1> (fx+ 10 1.0)
230436

in safe mode.  In unsafe mode the operation would not change at all.   
I suspect this is a much more controversial change than my last  
patch, but if felix and the community think it's OK, I'd be happy to  
edit the manual to reflect the new behavior.

Will

library.patch
Description: Binary data