On Thu, Sep 25, 2008 at 02:01:22PM -0700, chromatic wrote:
> On Thursday 25 September 2008 00:09:37 Moritz Lenz wrote:
> > As of today (and r31404) Rakudo's 'make spectest_regression' produces
> > compile time errors in three files, all of which are related to integer
> > overflow.
>
> I suspect it is too. I can't figure out what in PCT is the culprit
> (something related to printing floats),
Floats that are integer values greater than 2147483647 but less than
1e+15 don't have a decimal point or 'e' notation when they stringify,
so when IMCC gets a hold of them it carps about integers being out
of range.
Oddly, placing an 'L' at the end of the value seems to solve things:
assign $P0, 999999999999999 # IMCC integer overflow
assign $P0, 999999999999999L # works
I was working on having PCT recognize floats in this range and
then adding the 'L' when needed, but autopromoting I constants
to N constants seems like the right thing to do here.
More in a bit.
Pm
