Issue-1259 - HiGHS returns wrong solution

[erdembanak]

Scaled the residual for numerical tolerance.

Fix #1259

https://github.com/ERGO-Code/HiGHS/issues/1259

[erdembanak]

I have opened it since people might want to use it and it would be easier for you to control it. But why does infinity is causing a problem? I thought residual = std::max(residual, residual_rate) fallbacks to current implementation. (There might be other problems though)

[jajhall]

If either 'lower' or 'upper' is infinite then 'residual_range' is infinite. I don't know what 1/infinity gives with our infinity (kHighsInf). It might mean residual_rate is 0, so behaviour is unchanged, but I don't want it unguarded. I will look at this, don't worry!

[peno64]

Under Windows, 1.0 / Highs_getInfinity(highs) returns 0.0

[jajhall]

That's good to know, thanks