Formatted output with "%7.3dr" fails with "-0.0"

Users
1

It prints "+0.000"

Not urgent.

2

Hi Damian -

This thread seems more like a bug report. Anyway this worked for me:

var x = -0.0;

writeln(signbit(x));

writef("%7.3dr\n", x);

true
 -0.000

I see this on two different systems with or without --fast. So, could you make a bug report issue that also says more about your configuration?

1 Like
3

Sorry, only just noticed you had replied Michael. I will get onto it. Again, apologies.

4

Apologies, the problem was deeper than the formatting.

You were far wiser than me by checking the signbit beforehand.

The real culprit is the casting of a tuple to a complex.

Please try this.

// this is a real man's creation of a complex from its
// real and imaginary components, the latter as a real

proc cmplx(x : real(?w), y : real(w))
{
    var t : complex(w << 1);

    t.re = x;
    t.im = y;
    return t;
}


var x : complex(64) = (1.0:real(32), -0.0:real(32)):complex(64);
var y = cmplx(1.0:real(32), -0.0:real(32));

writeln("use standard Chapel casting of a tuple of 2 real to create complex");
writeln(signbit(x.re), " and ", signbit(x.im));
writef("%+7.3dr %+7.3dr i\n", x.re, x.im);

x = y;

writeln("use the 'cmplx' method to create complex from 2 individual reals");
writeln(signbit(x.re), " and ", signbit(x.im));
writef("%+7.3dr %+7.3dr i\n", x.re, x.im);

Have fun!

5

I am not seeing the problem. Your program has this output for me

use standard Chapel casting of a tuple of 2 real to create complex
false and true
 +1.000  -0.000 i
use the 'cmplx' method to create complex from 2 individual reals
false and true
 +1.000  -0.000 i

Is that not the output you expected? It seems to be what I would expect. Am I missing something?

1 Like
6

I see a plus sign on the 2nd line

using standard Chapel casting of a tuple of 2 real to create complex
false and false
+1.000 +0.000 i
using the 'cmplx' method to create complex from 2 individual reals
false and true
+1.000 -0.000 i

I am stuck with 1.22.0 because I cannot compile chapel 1.24.0 or above on my CentOS system.. And when I went digging into the compiler, I did not see changes. But I did not look where that casting was handled.

Sorry for the distraction.

We upgrade to CentOS 7.9 across the board so hopefully it has the tools, or I can add the tools, to handle the latest compiler.

7

FYI on older CentOS I've had success installing a newer GCC with a devtoolset package. See Developer Toolset 8 — Software Collections

I tried 1.22.1 on a recent Mac OS X system and still the program works as I showed before. So there may yet be something about your environment that causes the problem. I think the next step is to try it in your environment on the latest Chapel release.

1 Like
8

I managed to get some time on another machine with 1.22.0 under CentOS 7.

Probably does not exist in that environment. Sorry for the red-herring.