On Fri, 24 Aug 2012 12:20:58 -0700, Justin Lebar wrote:
>> inline int32_t GetFoo() {
>> int32_t result = 0;
>> nsresult rv = GetFoo(&result);
>> MOZ_ASSERT(NS_SUCCEEDED(rv));
>> return result;
>> }
>
> so that we never return uninitialized memory in release builds. 0 may
> or may not be right, but it should be no worse than a random value!
Initialization to a meaningless value is worse because it means
that valgrind (and perhaps other analysis tools) won't detect when
GetFoo is misbehaving.
_______________________________________________
dev-platform mailing list
dev-platform@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-platform
