Re: pdh: Remove or fix tests that fail on Windows.

[Hans Leidekker]

On Monday 22 October 2007 14:43:59 Jeff Latimer wrote:
> to test with and hence the delay.  The current patch I submitted was 
> cleaning up the Null reference issue before moving on to sorting out the

I don't see any crash on Wine or XP when running the test. You are
trying to protect against a hypothetical case where an app specifies
a size > 0 but no buffer. XP won't accept that anyway and it's unlikely
any app will do that on non-XP systems.

In fact your patch will break in another - also unlikely -
situation where an app relies on non-XP behaviour, which is to accept
a null buffer and size > 0 but less than the required size.

So a crash is even desirable here, because it tells us immediately
that the app is doing something peculiar.

 -Hans

pdh: Remove or fix tests that fail on Windows.

[Jeff Latimer]

Hans Leidekker wrote:
> On Monday 22 October 2007 14:43:59 Jeff Latimer wrote:
> > to test with and hence the delay.  The current patch I submitted was 
> > cleaning up the Null reference issue before moving on to sorting out the
>
> I don't see any crash on Wine or XP when running the test. 
> When I run
>     size = sizeof(buffer) / sizeof(WCHAR);
>     ret = PdhLookupPerfNameByIndexW( NULL, 674, NULL, &size );
> it crashes for me in wine and XP returns PDH_INVALID_ARGUMENT.
> You are
> trying to protect against a hypothetical case where an app specifies
> a size > 0 but no buffer. XP won't accept that anyway and it's unlikely
> any app will do that on non-XP systems.

I don't accept that it hypothetical because that is the exact behaviour 
the API spec encourages.  The buffer can be NULL in all circumstances.
> In fact your patch will break in another - also unlikely -
> situation where an app relies on non-XP behaviour, which is to accept
> a null buffer and size > 0 but less than the required size.

What is the result of that behaviour?  It looks to me that the only 
response should be PDH_INVALID_ARGUMENT or ok with an updated size.  As 
is stands pdh does neither.
> So a crash is even desirable here, because it tells us immediately
> that the app is doing something peculiar.

I don't understand this as there must be defined behaviour in this 
circumstance.

Jeff