[Ecls-list] read-char eof issue?

[Juan Jose Garcia-Ripoll]

On Sun, Aug 24, 2008 at 3:03 AM, Raymond Toy <toy.raymond at gmail.com> wrote:
> FWIW, I fixed the f2cl code and removed the recursive-p arg in several
> places.  (Don't know why it was set.)   With this change, ecl can now
> compile f2cl and can successfully convert all of the test programs that I've
> tried.  It works quite nicely now and compiles things quite fast and runs it
> quite fast too!

That's great news!

> So, if this is ancient ecl behavior, perhaps you should preserve it so you
> don't break other  existing code.

I do not think any other code relies on this "feature". I believe it
was more of an implicit assumption: if recursive-p is true then this
function is being called by ECL's reader and we want this or that
behavior.

But I am pretty sure that this old semantics is wrong, because, for
instance, :recursive-p = t  should be use when you implement a reader
macro and which is pretty much free to do whatever it wants with an
EOF. In particular, the macro may implement a comment that reads to
the end of file or until an closing character appears: that is
impossible if ECL aborts with both EOF and recursive-p = t.

So my only concern is whether other parts of ECL assume that old behavior.

Juanjo

-- 
Instituto de Física Fundamental
CSIC, Serrano, 113, Madrid 28040 (Spain)
http://juanjose.garciaripoll.googlepages.com