Re: [Axiom-developer] Axiom unsoundness

[Gabriel Dos Reis]

"Page, Bill" <address@hidden> writes:

| On Friday, October 27, 2006 1:01 PM Gabriel Dos Reis wrote:
| > 
| > Waldek Hebisch writes:
| > ... 
| > | 
| > |    >> System error:
| > |    Caught fatal error [memory may be damaged]
| > 
| > I've been running into this "memory may be damaged" stuff very
| > often these days ]with students, everything is possible :-)].
| > Most of them happens on invalid syntax and such.  The system
| > should be more resilient should not corrupt memory just because
| > of syntax errors and such.
| > 
| 
| Actually I believe that in *most* cases this error message
| greatly over states the case and is probably rather poorly
| choosen from a user psychology point of view. (Many of Axiom
| error message suffer from this.) In spite of this message,
| have not seen  any clear evidence of corrupt memory. Really
| this is a Lisp issue and my experience with GCL is that in
| spite of some surprising things that it does to manage
| memory, it does in fact do quite a good job.
| 
| I'd be interested if you have any evidence actual memory
| problems.

Here is a case where the system did not issue a diagnostic, but just
terminates with exit status 0 (!) [so it is not a case of message with
evidence of actual memory damage.  It is a case of non-message with
actuall memory damage.].  Type in

     VALUES(Integer)$Lisp

Even when we think the operation is meaningless, the system should try
hard not to crash.  In fact, the system can predict that the operation
is nonsensical and issue appropriate message.

-- Gaby