[Bug c++/110749] New: wrong show float

[Bug c++/110749] New: wrong show float

[alexsyrezerv at mail dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110749

            Bug ID: 110749
           Summary: wrong show float
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: alexsyrezerv at mail dot ru
  Target Milestone: ---

Created attachment 55586
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55586&action=edit
To try on different compillers

Code:
```
#include <iostream>
#include <iomanip>

void func_float_point(std::ostream& os, const double& value, const std::size_t&
size) {
  os << (0.0 > value ? '-' : '0') << std::setw(size - 1) << std::setfill('0')
<< std::abs(value);
}

int main(int, char**) {
  const double a[] = {90.8425, 101.9225};
  for(std::size_t i = 0; i < 2; ++i){
    func_float_point(std::cout, a[i], 17);
    std::cout << std::endl;
  }

  return 0;
}
```
result
```
butiaev@butiaev-VirtualBox:~/tmp$ make bug
g++     bug.cpp   -o bug
butiaev@butiaev-VirtualBox:~/tmp$ ./bug
000000000090.8425
0000000000101.922
butiaev@butiaev-VirtualBox:~/tmp$

```

[Bug c++/110749] wrong show float

[schwab@linux-m68k.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110749

Andreas Schwab <schwab@linux-m68k.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |INVALID
             Status|UNCONFIRMED                 |RESOLVED

--- Comment #1 from Andreas Schwab <schwab@linux-m68k.org> ---
101.9225 cannot be represented exactly, the nearest representable value is
101.92249999999999943.

[Bug c++/110749] wrong show float

[alexsyrezerv at mail dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110749

--- Comment #2 from Aleksei <alexsyrezerv at mail dot ru> ---
Created attachment 55589
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=55589&action=edit
GDB SHOW IT

But GDB show it normally

[Bug c++/110749] wrong show float

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110749

--- Comment #3 from Jonathan Wakely <redi at gcc dot gnu.org> ---
Your program is using the default precision for printing a double: 6
significant figures.

101.92249999999999943 with 6 digits of precision is 101.922 (because (24 rounds
to 2).

101.92249999999999943 with 7 digits of precision is 101.9225 (because 49 rounds
to 5).

If you use std::setprecision(7) << std::abs(value) you will get 101.9225:

000000000090.8425
000000000101.9225

If you use std::setprecision(17) you will get the same values as GDB shows:

090.842500000000001
000000000101.9225

If you use std::setprecision(18) you'll see that Andreas is correct:

090.8425000000000011
0101.922499999999999


This is not a GCC bug, this is just how floating-point numbers work.

[Bug c++/110749] wrong show float

[alexsyrezerv at mail dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110749

--- Comment #4 from Aleksei <alexsyrezerv at mail dot ru> ---
ок

[Bug c++/110749] wrong show float

[alexsyrezerv at mail dot ru]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110749

--- Comment #5 from Aleksei <alexsyrezerv at mail dot ru> ---
this variant more clear.. 4 symbol after point// may by c++ has modification
for set symbols after decimal point ? 
```
  void func_float_point(std::ostream& os, const double& value, const
mcx::convert_info_t& cv) {
    char buf[buf_size];
    snprintf(buf, buf_size - 1, "%0*.*f", cv.size, cv.pos, value);
    os << buf;
  }
```

[Bug c++/110749] wrong show float

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110749

--- Comment #6 from Jonathan Wakely <redi at gcc dot gnu.org> ---
If you use std::fixed as the floatfield format then std::setprecision specifies
the number of digits after the decimal point:

os << (0.0 > value ? '-' : '0') << std::setw(cv.size - 1) << std::setfill('0')
<< std::setprecision(cv.pos) << std::fixed << std::abs(value);


You don't need to handle the -'- sign yourself:

os << std::setw(cv.size) << std::internal << std::setfill('0') << std::fixed <<
std::setprecision(cv.pos) << value;


This matches the output of printf("%0*.*f", cv.size, cv.pos, value)

[Bug c++/110749] wrong show float

[redi at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=110749

--- Comment #7 from Jonathan Wakely <redi at gcc dot gnu.org> ---
(In reply to Jonathan Wakely from comment #6)
> You don't need to handle the -'- sign yourself:

Oops, I mean '-' sign.