This is the mail archive of the guile@sourceware.cygnus.com mailing list for the Guile project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Regarding recent void * change


> Pairs, vectors, objects etc now consist of fields of the type void *.
> When dealing with such values, the compiler should be able to make the
> assumption that we're dealing with pointers, so it can use registers
> in the CPU specialized for holding addresses.
> 
> I'm afraid that there might exist architectures where dealing with
> values which are invalid as pointers will yield a bus error.

This is a rather vague potentiality. I can think of early 68k machines
where bus errors resulted from dereferenceing a pointer to an odd address
but never from storing an odd number in an address register. Maybe it is
possible to find an architecture whose address registers are so specialised
that they cannot cope with some numbers but that would be a very unusual
case and should not distort the core effort. The magic number for human
perception is 95%, once something is true 95% of the time, it is true.

> I think we generally should think more about how changes influence
> Guile on other architectures than we have done recently.

Don't go looking for outlandish cases though. There are enough problems
to solve right here and now without finding problems that may sometime
happen perhaps.

	- Tel


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]