Feature request

[v3cz0r]

So what happened to your data-flow analysis? Or perhaps, there's none, just a dead simple check?

Oh, ok, you were just trying to bash FB as always, no problem then, i stopped taking anything you say into account since ages ago.

[marzecTM]

but meh, fb sucks anyways :p SCNR

learn to read and interpret properly. anyways, i still don't see where the problem is with detecting something like the above. there's of course no need for fullfledged data flow analysis, it's actually a "dead" simple pass to the type checker. nothing more nothing less, never said i implemented it either.

this is all just something we do in our free time, aka hobby. don't get to serious about it :p

[LooseCaboose]

anyways, i still don't see where the problem is with detecting something like the above. there's of course no need for fullfledged data flow analysis, it's actually a "dead" simple pass to the type checker.

It can be done for simple cases and is, as you say, dead easy. However in the general case the problem is undecideable; it cannot be done no matter how much dataflow analysis etc is done. This is why gcc and vc++ don't report errors for the code I gave.

In my example you cannot warn that the function returns a pointer to a local variable since it doesn't always. Even warning about possibly returning a pointer to a local variable may be over zealous since the callee may never give values of x and y which allow it to happen.