[1/2] shell32/tests: Trace the last ShellExecute command whenever a corresponding test fails.

[Francois Gouget]

On Fri, 22 Jan 2016, Alexandre Julliard wrote:

> Francois Gouget <fgouget at codeweavers.com> writes:
> 
> > On Fri, 22 Jan 2016, Alexandre Julliard wrote:
> >> OTOH, it means that in silent mode you get no indication at all.
> >
> > ??? That's what we want right?
> >
> > We don't want tests that succeed to spew out thousands of lines of 
> > traces because that would just bloat the logs, causing them to be 
> > rejected by test.winehq.org.
> 
> On failure you should be able to see what caused the failure, even in
> silent mode. With your change you'd have to re-run it with debug on, and
> no guarantee that you can get the same failure to happen again.

Oh, I see, in the WINETEST_DEBUG=0 case winetest_trace() does not print 
anything. I never run in that mode (bug the testbot and WineTest do) so 
I missed that. I will send an updated patch.


[...]
> > Besides, even if we never add any new test we will get failures. All it 
> > takes is a new Windows version like Windows 10.
> 
> Of course, I'm not arguing for not logging failures. On the contrary, my
> argument is that there's no reason to go out of your way to reduce
> output on failure (as opposed to output on success, which I agree should
> be minimized).

It also makes the test failures more readable and has the benefit of 
encapsulating shell_call which is much better code-wise.


-- 
Francois Gouget <fgouget at codeweavers.com>