Re: errors and Side-Effect-class

[Bruno H. <br...@cl...>]

Sam wrote:
> > Not all functions, but those which have a non-empty "Exceptional Situations"
> > section in CLHS yes.
> 
> are you saying that a function for which CLHS does not prescribe an 
> "Exceptional Situation" (e.g., a function not mentioned there due to being a 
> CLISP exception) can be folded even if clisp may signal an error in it?
> this - at the very least - introduces differences between compiled and 
> interpreted code.

Yes. And this is where the OPTIMIZE quality comes into play. If someone is
compiling with maximum SAFETY, keep the call that potentially could signal
a condition even if CLHS does not require it. If someone is compiling with
SAFETY 0, and CLHS does not require the error checking, drop it.

> > The seclass to be used for them is - see for example in c-THROW - an seclass
> > with :modifies 'T. The :uses and :uses-binding parts can be left unchanged.
> 
> well, this basically means that using seclass_default for all subrs which 
> signal errors.

Have you checked this in CLHS? You need to check them one by one, IMO.

Bruno