Re: [Seed7-users] Seed7 Compilation Errors

[Thomas M. <tho...@gm...>]

Hi Zachary,

Sorry for the delay, I was rather busy with graphic formats.

You write:
> Hey Thomas,
>
> Unfortunately I am still getting errors. However, this time I modified the
> chkflt.sd7 program to output the result of the trunc() and round() operations
> I'm not sure why I  didn't think of this before.
> I've attached the files so you can see for yourself, but it appears that
> truncating or rounding 9223372036854775807.0 results in -9223372036854775808.
> Which you did mention was a possibility in your last e-mail.
>
> Does this mean something like
> "and trunc(9223372036854775807.0) <> -9223372036854775808" is needed?

No, it probably means that MAXIMUM_TRUNC_ARGUMENT has the wrong value.
For some reason chkccomp.c does not compute MAXIMUM_TRUNC_ARGUMENT correct on
your computer.

I created a test program:
========== begin chktrunc.c ==========
#include "stdio.h"
#include "float.h"
#include "version.h"
#include "common.h"

double maxFltValues[] = { -9223372036854773760.0, -9223372036854774784.0,
                          -9223372036854775808.0, -9223372036854777856.0,
                           9223372036854773760.0,  9223372036854774784.0,
                           9223372036854775808.0,  9223372036854777856.0 };
int64Type maxIntValues[]  = { -9223372036854773760,   -9223372036854774784,
                              -9223372036854775807-1, -1234567890123456789,
                               9223372036854773760,    9223372036854774784,
                               9223372036854775807,    1234567890123456789};

int main (int argc, char *argv[]) {
  int pos;
  char *outcome[] = {"fail   ", "success"};
  int64Type minimumTruncArgument;
  int64Type maximumTruncArgument;
  for (pos = 0; pos < 8; pos++) {
    printf("%f: " FMT_D " %s " FMT_D " %f %d %d %d %d\n",
           maxFltValues[pos], maxIntValues[pos],
           outcome[(int64Type) maxFltValues[pos] == maxIntValues[pos]],
           (int64Type) maxFltValues[pos], (double) maxIntValues[pos],
           maxFltValues[pos] < (double) -9223372036854774784,
           maxFltValues[pos] > (double) 9223372036854774784,
           maxFltValues[pos] < (double) -9223372036854775807-1,
           maxFltValues[pos] > (double) 9223372036854775807);
  }
  for (pos = 0; pos < 4; pos++) {
    if ((int64Type) maxFltValues[pos] == maxIntValues[pos]) {
      minimumTruncArgument = maxIntValues[pos];
    }
  }
  for (pos = 4; pos < 8; pos++) {
    if ((int64Type) maxFltValues[pos] == maxIntValues[pos]) {
      maximumTruncArgument = maxIntValues[pos];
    }
  }
  printf("#define MINIMUM_TRUNC_ARGUMENT " FMT_D "\n", minimumTruncArgument);
  printf("#define MAXIMUM_TRUNC_ARGUMENT " FMT_D "\n", maximumTruncArgument);
  return 0;
}
========== end chktrunc.c ==========
This program includes version.h and common.h. You can it compile and run with:
  gcc chktrunc.c -o chktrunc
  ./chktrunc
On my computer this results in:
-9223372036854773760.000000: -9223372036854773760 success -9223372036854773760 -9223372036854773760.000000 0 0 0 0
-9223372036854774784.000000: -9223372036854774784 success -9223372036854774784 -9223372036854774784.000000 0 0 0 0
-9223372036854775808.000000: -9223372036854775808 success -9223372036854775808 -9223372036854775808.000000 1 0 0 0
-9223372036854777856.000000: -1234567890123456789 fail    -9223372036854775808 -1234567890123456768.000000 1 0 1 0
9223372036854773760.000000: 9223372036854773760 success 9223372036854773760 9223372036854773760.000000 0 0 0 0
9223372036854774784.000000: 9223372036854774784 success 9223372036854774784 9223372036854774784.000000 0 0 0 0
9223372036854775808.000000: 9223372036854775807 fail    -9223372036854775808 9223372036854775808.000000 0 1 0 0
9223372036854777856.000000: 1234567890123456789 fail    -9223372036854775808 1234567890123456768.000000 0 1 0 1
#define MINIMUM_TRUNC_ARGUMENT -9223372036854775808
#define MAXIMUM_TRUNC_ARGUMENT 9223372036854774784

Part of chktrunc.c is code that I use in an improved chkccomp.c.
Please run it on your computer and tell me the results.

Regards
Thomas