[haiku-commits] Re: r38135 - haiku/trunk/src/preferences/print

  • From: Michael Pfeiffer <michael.w.pfeiffer@xxxxxxxxx>
  • To: haiku-commits@xxxxxxxxxxxxx
  • Date: Mon, 16 Aug 2010 18:52:51 +0200

Am 16.08.2010 um 17:00 schrieb pulkomandy:

> On Mon, Aug 16, 2010 at 04:53:20PM +0200, Michael Pfeiffer wrote:

> static_cast will never return NULL.

That's the problem and may hide a future bug if the code ever changes.

> But you have to be sure there are
> only PrinterItem in the list or the object returned will be invalid, or
> there will be a runtime exception. In this case I don't think this will
> ever happen unless someone really mess with the code, so the check isn't
> useful.

You cannot know that it won't ever change. It is much safer to have runtime 
checks in source code than to rely on the knowledge of a human about the code.

> Using dynamic_cast is not useful in these cases and is also slower.

Being slower does not matter in this case as this is no time critical code.

> And doing an assert would be a bit silly. If the only way to avoid a crash
> is to crash before it, why bother adding code for that ?

Using an assertion ensures the crash to happen near to cause of the problem. I 
wouldn't call that silly.

Actually, I don't know if using a NULL pointer leads always immediately to a 
crash. When I wrote the code I assumed that that would happen. But since 
Coverty complains about it there should be a solution applied that does not 
make the code unsafer.

Any other opinions?


Other related posts: