#3880 expr returns incorrect results on M68K architecture

[Sergei Golovan]

Hi!

expr in current HEAD returns the following on Motorola M68K architecture:

% expr 0.1
0.000
% expr 1e-10
0e-12

The bug is triggered only when the final result is shown, and only when the resulting number is less than one in absolute value. Then expr replaces mantissa by zero.

% expr (1.0/10)
0.000
% expr (1.0/10)*10
1.0

Also, expr.test fails with the following:
expr.test
Test file error: unable to alloc 3988184000 bytes

(I don't know if it's the same bug or another one).

8.4.16 works fine and passes all the tests.

[Don Porter]

Logged In: YES
user_id=80530
Originator: NO

Is this another symptom of 1851832?

[Nobody/Anonymous]

Logged In: NO

Even if it's a symptom of 1851832 the current HEAD (with a patch from 1851832) doesn't cure it.

[miguel sofer]

Logged In: YES
user_id=148712
Originator: NO

I think Tcl_Objs are properly aligned, my guess is that the problem is in TclDoubleDigits(). I will not be able to look at this in the next few days.

Please report more platform details - is this linux? What libc?

[Sergei Golovan]

Logged In: YES
user_id=410366
Originator: YES

Yes, it's Debian Linux, glibc version is 2.5.

[Don Porter]

Logged In: YES
user_id=80530
Originator: NO

Wonder if this is an issue
analogous to what we faced
with the Nokia 770? Yet
another wacky endian variation?

http://wiki.tcl.tk/15408

[Sergei Golovan]

Logged In: YES
user_id=410366
Originator: YES

It's a bug in frexp() function. The following program:

#include <math.h>
#include <stdio.h>

main() {
double v;
int e;
v = frexp(1.1, &e);
printf("%g, %d\n", v, e);
v = frexp(0.1, &e);
printf("%g, %d\n", v, e);
}

prints out
0.55, 1
0.05, -2147483647

which is obviously wrong. The result should be
0.55, 1
0.8, -3

So, there's no bug in Tcl itself.