Re: More safety for "space leak fix"

>>>>> "Stephen" == Stephen J Turnbull <[email protected]> writes:

>>>>> "ms" == Michael Sperber <[email protected]> writes:

ms> Essentially, I had forgotten about
ms> `dontusethis-set-symbol-value-handler.'  (Primarily because I
ms> hadn't anticipated that crapola like this would get into the
ms> system,

Stephen> Uh, what's it allegedly good for?  And maybe it should come back out?
Stephen> (Please push that discussion to -design if you have a reason to
Stephen> actually answer those questions.  :-)

It allows you to register a callback for a variable that gets invoked
when someone tries to set that variable.  Ben used this to preservew
some historical aspects of the API.  (Mostly in Mule, which is why I
suspect I wasn't seeing the freed lrecord bug.)

Hrvoje is right that we need it for preserving the API.  On the other
hand, it completely violates some pretty basic semantic assumptions
you might mentally make when looking at code.  That what happened to
me: I thought Fset couldn't *possibly* GC.  Which is why I think
changing those APIs, with all the costs that entails, would have been
preferable.

I vaguely do remember discussion about this.  I also vaguely remember
there being some pretty strong opposition.  (I know there was from
me.)  But Martin and Ben seem to've put it in anyway.

-- 
Cheers =8-} Mike
Friede, Völkerverständigung und überhaupt blabla