x86_64 TCG: Incorrect floating point cast to int.

[Vasily Dybala]

I used exaample from:
https://stackoverflow.com/questions/3986795/what-is-the-result-of-casting-float-inf-inf-and-nan-to-integer-in-c

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

int main(int argc, char** argv) {
  float a = INFINITY;
  float b = -INFINITY;
  float c = NAN;

  printf("float %f %f %f\n", a, b, c);
  printf("int %d %d %d\n", (int) a, (int) b, (int) c);
  printf("uint %u %u %u\n", (unsigned int) a, (unsigned int) b, (unsigned int) c);
  printf("lint %ld %ld %ld\n", (long int) a, (long int) b, (long int) b);
  printf("luint %lu %lu %lu\n", (unsigned long int) a, (unsigned long int) b, (unsigned long int) c);

  return 0;
}

And got different results on real computer and on qemu.

output from real HW is the same as on stackoverflow:

$ gcc test.c && ./a.out
float inf -inf nan
int -2147483648 -2147483648 -2147483648
uint 0 0 0
lint -9223372036854775808 -9223372036854775808 -9223372036854775808
luint 0 9223372036854775808 9223372036854775808

But on qemu I got another results:

float inf -inf nan
int 2147483647 -2147483648 2147483647
uint 4294967295 0 4294967295
lint 9223372036854775807 -9223372036854775808 -9223372036854775808
luint 18446744073709551615 9223372036854775808 9223372036854775807

qemu launch string:
/qemu-system-x86_64 -m 1024 -cpu core2duo -serial stdio -netdev user,id=network0 -device e1000,netdev=network0 -kernel my_kernel

qemu version:
x86_64-softmmu/qemu-system-x86_64 --version
QEMU emulator version 3.1.50 (v3.1.0-1676-ge47f81b617-dirty)
Copyright (c) 2003-2019 Fabrice Bellard and the QEMU Project developers

This bug affect some javascript (surprise) calculations:

var conversion = "01234567890";
var x;
var result = conversion[x & 42];
console.log(result)

In example, var x is "undefined"
and when do calculation "x & 42" on js we should get 0 (it is documented feature), but actually got "42"

and "result" sould be "0" but actually we got "undefined"

[Peter Maydell (pmaydell)]

https://<email address hidden>/ is a patch which fixes the C test case (and may also fix the node.js case, though I don't have a setup to test that).

[Peter Maydell (pmaydell)]

This should be fixed by commit 1e8a98b53867f61da9, which will be in the 4.2 release.