#2214 FreeBSD, expr, "Infinity"

[Anonymous]

From my testing, works on FreeBSD, but not on
NetBSD. Tested on 8.3.2, then upgraded to 8.3.5
(same issue).

-> expr {0?1:"Infinit"}
<- Infinit

-> expr {0?1:"Infinity"}
<- Tcl error: floating-point value too large to represent

It's being messed up by the word 'infinity'?

-----------

OS Platform and Version
-- FreeBSD 1.6E

Problem Behavior
-- The word 'Infinaty' not liked in expr evaluations?

Expected Behavior
-> expr {0?1:"Infinity"}
<- Infinity

-----------

Carl M. Gregory
http://mc.purehype.net/whois/

[Carl M. Gregory]

Logged In: YES
user_id=721653

OS Platform and Version
-- NetBSD 1.6E

Sorry

[Donal K. Fellows]

Logged In: YES
user_id=79902

Bet this is yet another native strtod() problem. :^(

[Don Porter]

Logged In: YES
user_id=80530

Note related change in Tcl 8.4:

2002-07-22 Donal K. Fellows <fellowsd@cs.man.ac.uk>

* tests/expr.test (expr-22.*): Added tests to help
detect the
corrected handling.
* generic/tclExecute.c (IllegalExprOperandType):
Improved error
message generated when attempting to manipulate Inf
and NaN values.
* generic/tclParseExpr.c (GetLexeme): Allowed parser
to recognise
'Inf' as a floating-point number. [Bug 218000]

[Jeffrey Hobbs]

Logged In: YES
user_id=72656

Kevin's playing with this now...

[Kevin B KENNY]

Logged In: YES
user_id=99768

Infinity and NaN are now recognized as floating-point "number"s
on all platforms with IEEE-754 arithmetic. Infinity does
not cause an error to be thrown, and tests larger than any
finite number. NaN causes an error to be throws whenever
it arrives in Tcl_GetDoubleFromObj. It is handled silently
in [binary format] and [binary scan].

"Fixed in 8.5"