Tcl Source Code

I'm checking input values and see this strange
behaviour of the 'string is integer' command.

puts "tcl_version: $tcl_version"
set numbers {1 +1 ++1 +-1 -+1 -1 --1 "-  +1"
                   "+  \n+1" "-  \t-1"}
foreach num $numbers {
  puts "$num is integer: [string is integer $num]"
}

See the results of '[string is real]':

set numbers {1.0 +1.0 ++1.0 +-1.0 -+1.0 -1.0 --1.0 "- 
+1.0" 
"+  \n+1.0" "-  \t-1.0"}
foreach num $numbers {
   puts "$num is double: [string is double $num]"
}

Logged In: YES 
user_id=72656

corrected for 8.4.2.

File Added - 35424: string634856.patch

Logged In: YES 
user_id=72656

Attached is a patch which corrects this behavior in the head.  
I completely removed the funky functionality with #ifdef 
because I couldn't find a system to test a "bad strtoul" 
implementation.  According to the C standard and reading of 
libc implementations, we shouldn't ever have this problem, but 
I left in at least the #ifdef in case someone finds such a 
system.

Logged In: YES 
user_id=99768

The problem is actually in Tcl_GetIntFromObj, as the following shows.
Note that literal sharing confuses the issue, since once the object
acquires the integer internal representation that it shouldn't have, it
keeps that internal rep.

(bin) 49 % info patchlevel
8.4.0
(bin) 50 % set x "+ -1"
+ -1
(bin) 51 % expr { $x + 0 }
can't use non-numeric string as operand of "+"
(bin) 52 % set y [format %d $x]
-1
(bin) 53 % set x
+ -1
(bin) 54 % expr { $x + 0 }
-1