Re: Bug with *print-base* 36

On Sun, 8 Jun 2003, Patrik Nordebo wrote:

> While porting Gilbert Baumann's web browser closure to OpenMCL, I ran
> into what looks like a bug in the printer when *print-base* is bound
> to 36 and a bignum is printed. The bug is in ccl::%pr-integer, but the
> code looked too hairy for me to fix. Below is a transcript that
> demonstrates the problem.
>
> Welcome to OpenMCL Version (Beta: Darwin) 0.13.5!
> ?
> ? (setq *print-base* 36)
> 10
> ? (expt 36 7)
> > Error: Array index Z out of bounds for #<SIMPLE-VECTOR 35> .
> > While executing: CCL::%PR-INTEGER
> > Type :POP to abort.
> Type :? for other options.
> 1 > :pop
> ? (setq *print-base* 35)
> 10
> ? (expt 36 7)
> 17M10L71
> ? (1- (expt 35 7))
> YYYYYYY
> ? (ccl::%pr-integer (expt 36 7) 36)
> > Error: Array index 35 out of bounds for #<SIMPLE-VECTOR 35> .
> > While executing: CCL::%PR-INTEGER
> > Type :POP to abort.
> Type :? for other options.
> 1 >
>

The vector that's one element too small is hardwired into the
definition of %PR-INTEGER itself; it's used to estimate the number of
digits (in the current radix) that'll be needed to print a 32-bit
chunk of a bignum.  It's the initial value of the variable SIZE-VECT,
and I have no idea why it's defined with #.(vector ...) instead of
with #(...); I have a hunch that this code was written a long, long
time ago.

Adding another "7" at the end of the vector would work (if you could
print a 32-bit integer in 7 base-35 digits, it can't require more digits
in base 36).

If you try to fix this by interactively redefining %PR-INTEGER, note
that it depends on macros that're defined in files that're ordinarily
REQUIREd during compilation (see the EVAL-WHEN near the top of
"ccl:level-0;l0-int.lisp".)

Thanks.