[Openmcl-devel] Re: *save-definitions*

Gary Byers gb at clozure.com
Wed Jun 2 23:21:54 UTC 2004



On Wed, 2 Jun 2004, Hamilton Link wrote:

> It strikes me that *save-definitions* could default to T since memory
> is cheap.
>

I don't have a good sense of what the space (and GC) impact of this
would be.

I can certainly imagine cases where saving bits and pieces of the
original lambda expression could be very useful (e.g., mapping PC
ranges in compiled code to "the source").  It's not clear that saving
the whole lambda expression is too useful in and of itself; it allows
FUNCTION-LAMBDA-EXPRESSION to return more useful/accurate information
(and helps functions like ARGLIST be more accurate in a few cases).
Is this compelling ?

> And one possible deviation from ANSI might be good, if in fact it's a
> deviation at all: I would recommend *read-eval* default to NIL (even if
> the openmcl sources need it on, it's easy enough to turn it off in
> save-app and turn it back on in xload-level-0 etc.). The people whose
> code it breaks can turn it on easily (if you use #. you should know the
> variable exists), but it does represent a security hole in everyone's
> apps that do a READ on a TCP stream.
>

I don't think that breaking #. would be wise, though it might be wise
to stress the fact that reading S-expressions from untrusted sources
can have unexpected side-effects.

> h
>
>
>



More information about the Openmcl-devel mailing list