advapi32: Fix potential NULL pointer dereference in RegSetValueExA [with test] (Saturn)

[Aurimas Fišeras]

Alexandre Julliard wrote:
> Rob Shearman <robertshearman at gmail.com> writes:
> 
>> 2009/1/26 Aurimas Fišeras <aurimas at gmail.com>:
>>> Saturn's error report:
>>> (INCONSISTENT USE) Possible null dereference of variable data+(count-1).
>>> This variable is checked for Null at lines: registry.c:1051
>>>
>>> Tested on Windows XP
>>>
>>> Changelog:
>>>    advapi32: Fix potential NULL pointer dereference in RegSetValueExA
>>> [with test] (Saturn)
>> Excellent, this tool has spotted a corner-case that the code doesn't
>> handle correctly.
> 
> I'm not convinced that this is really a bug. If a non-zero count is
> specified it's quite reasonable to expect data to be valid.

It is also quite reasonable to expect that a function won't crash with
all legal parameters.

> 
> Of course Windows has exception handlers all over the place, but that
> doesn't mean we want to replicate that behavior.

But we want to have a "bug-for-bug" compatibility with Windows?
Without this patch windows just returns an error, while wine crashes.

There are dozens of similar corner-case errors where Windows crashes as
well as wine, but this time only wine crashes.