Re: shell32/tests: Arrange for the shell_execute{_ex}() calls to set the test line.

[Alexandre Julliard]

Francois Gouget  writes:
> This makes it possible to attribute their internal test results to the relevant test_xxx() function.
> This implies that they must not make calls that would reset the test line.

This is fragile. It would be better to explicitly pass the line number,
like most other tests do.

shell32/tests: Arrange for the shell_execute{_ex}() calls to set the test line.

[Francois Gouget]

On Fri, 15 Jan 2016, Alexandre Julliard wrote:
> Francois Gouget  writes:
> > This makes it possible to attribute their internal test results to the relevant test_xxx() function.
> > This implies that they must not make calls that would reset the test line.
>
> This is fragile.

How so? This is how the ok() calls work.
> It would be better to explicitly pass the line number, like most other 
> tests do.

So I guess you object to the okChildXxx() patch too.

The drawback of explicitly passing the file/line number is that it adds 
hidden parameters which makes interpreting compiler errors more complex. 
It also means using two different mechanisms instead of one which also 
increases complexity.

shell32/tests: Arrange for the shell_execute{_ex}() calls to set the test line.

[Alexandre Julliard]

Francois Gouget  writes:
> On Fri, 15 Jan 2016, Alexandre Julliard wrote:
> > Francois Gouget  writes:
> > > This makes it possible to attribute their internal test results to the relevant test_xxx() function.
> > > This implies that they must not make calls that would reset the test line.
> >
> > This is fragile.
>
> How so? This is how the ok() calls work.

winetest_set_location should be used only for a short amount of time,
like inside the ok() call. You can't assume that it's going to be
unchanged across a large function. Nobody is going to remember that this
function is special and can't use ok() like everywhere else.