Hi Cyril,
Thanks for this, however there remains one (small) corner case : if dtmout
is equal to INT_MAX (2147483647), it will be rounded up by the +0.5 then
become negative :
On Sun, Aug 19, 2018 at 12:27:13AM +0200, Cyril Bonté wrote:
> if (dtmout > INT_MAX) /* overflow check */
> - WILL_LJMP(luaL_error(L, "settimeout: cannot set values larger
> than %d", INT_MAX));
> + WILL_LJMP(luaL_error(L, "settimeout: cannot set values larger
> than %d ms", INT_MAX));
Above it will pass the test.
> - tmout = MS_TO_TICKS((int)dtmout);
> + /* round up the timeout to convert it to int */
> + tmout = MS_TO_TICKS((int)(dtmout + 0.5));
and here we convert 2147483647.5 to int, which becomes -2147483648.
Given that the double to int conversion already rounds values, we do not
need the +0.5 so we can get rid of it. Alternatively, dtmout could be
rounded up before the checks using round() but apparently this is not
needed.
I suspect that the +0.5 might have been placed here to ensure that no
short timeout is rounded down to zero. If that's the case we could
simply proceed like this :
if (dtmout < 0)
...
if (dtmout > INT_MAX)
...
tmout = MS_TO_TICKS((int)dtmout);
if (tmout == 0)
tmout++; // min 1ms
Just let me know if you want me to adjust the patch one way or another
so that you don't need to respin one.
Thanks!
Willy
