WTG3012 needs to restore leading trivia

[andrewhongnsw]

In the following situations, WTG3012 removes the first preprocessor directive entirely, leading to CS1027 errors:

Example 1

return
#if DEBUG
     something.HasValue ? something.Value :
#endif
true;

Example 2

return false
#if DEBUG
|| (
    value0 &&
    value1 &&
    value2
)
#endif
;

Example 3

protected virtual bool ShouldDoSomething => true
#if DEBUG
    && !(value0 && value1)
#endif
;

[brian-reichle]

WTG3012 doesn't consider the pre-processor directives, maybe we could simply not offer a code-fix if the expression contains pre-processor directives; but all those examples represent dodgy practices that would fail on at least two other rules. I'm not sure if this is a defect worth fixing.

to me, they would be better written as (assuming the condition on DEBUG is unavoidable):

#if DEBUG
if (something.HasValue)
{
    return something.Value;
}
#endif
return true;

#if DEBUG
if (value0 && value1 && value2)
{
    return true;
}
#endif
return false;

#if DEBUG
protected virtual bool ShouldDoSomething => !(value0 && value1);
#else
protected virtual bool ShouldDoSomething => true;
#endif

[yaakov-h]

All three of these examples already trigger WTG3103 (and WTG2002, but that's due to the name of the symbol).

That rule has been suppressed in a few repos due to a pre-existing habit of doing things like this.

I'd be tempted to leave it as is, but I think the more comprehensive approach is to suppress the Diagnostic (or just suppress the code-fix) if the expression contains preprocessor directives, and rely on WTG3103 to catch cases like this.

If a team then turns off WTG3103, they only have themselves to blame for tricky code like this.