On Thu, 20 Sep 2007, Maciej W. Rozycki wrote:

 Perhaps preinitialising to an error value such as -EINVAL would be of
more sense.  This way any error paths lacking initialisation are still
reported as errors, even though the classification might be wrong.

Eeee ... at least I wouldn't prefer that. Why not simply use the
"int x = x;" trick (which is what uninitialized_var() does) -- it shuts
up the warning, and does *nothing* else. The bug will not be hidden, if
there's bad misbehaviour happening due to the bug, it will continue to
happen that way -- thus bringing our attention to it. Pre-initializing
to -EINVAL (or whatever) has the problem that when the bug actually
triggers, something unrelated might happen higher up the callchain, and
we'd be scratching our heads in a "why are we getting a -EINVAL here?"
kind of way ... worse still, we might think that this was _really_ an
EINVAL and go about debugging it ...

Plus, pre-initializing to -EINVAL (or even 0) will waste some bytes in
kernel text size, but no such overhead with uninitialized_var() :-)


Satyam