[PATCH] 2/3 mtd: add support for flash on the SEGA Dreamcast Visual Memory Unit

[Andrew Morton]

On Mon, 24 Mar 2008 15:54:33 +0100
J__rn Engel <joern at logfs.org> wrote:

> On Mon, 24 March 2008 14:43:16 +0000, Adrian McMenamin wrote:
> > 
> > Well, I haven't got round to applying them and testing them yet, though
> > they look ok, obviously, except for one thing: I'm pretty much where you
> > are on the goto versus return thing, but got pretty clear instructions
> > on a previous patch from Andrew Morton that using gotos to ensure
> > functions limit the number of return points is the way to go.
> > 
> > I've added him back in now we aren't exchanging flames, so maybe he can
> > pronouce ex cathedra.
> 
> It is a matter of personal taste.  Having a single return statement is
> nice when using a debugger.  One can set a single breakpoint instead of
> ten.  I guess that is why Andrew prefers it.

Experience has shown that over time, the multiple-return-point approach
leads to locking errors, resource leaks and to much duplicated
unlocking/freeing code.

Because when people add new locking and more dynamic allocations to the function
they need to hunt down each `return' and add unlocking/freeing code to it (this
is bad).   And sometimes they miss one (this is worse).