On Jan. 6, 2012, Raymond Toy wrote:
-------------------------------
>This is really an issue with gcl. It looks like gcl doesn't signal an
>errors on overflow and uses infinities instead. Other lisps (cmucl and
>ccl) do signal errors which are silently caught and a noun form is
>returned.
>
>I suppose the routines could check that no infinities or NaNs are returned
>in the result.
-------------------------------
Thanks for the info. I have inserted an additional check in parts of
nint.mac, such as
------------------------
rL : apply ('quad_qagi, argL),
if (substring(string(rL[1]),1,8) = "1..#INF" ) then
(print(" quad_qagi: overflow"),
ans : false)
etc.,
-----------------
which is no big deal. If some of the experts would post
Windows binaries using other lisps (cmucl, ccl,..) it would
make it easier for Windows slaves (like myself) to check
code construction problems of this type.
Ted