[PATCH v4 07/10] shell32/autocomplete: Redesign the window proc to trigger on key presses instead of key release

[Gabriel Ivăncescu]

On Tue, Sep 11, 2018 at 9:51 PM, Henri Verbeet <hverbeet at gmail.com> wrote:
> On 11 September 2018 at 23:05, Huw Davies <huw at codeweavers.com> wrote:
>> It makes following the flow of the code incredibly difficult.  If you
>> don't believe me, come back to this patch in a month and see how you feel
>> then.
>>
> Yeah, this. You have goto's in fairly deeply branched code, inside a
> nested switch, jumping to a branch somewhere inside the outer switch,
> all of which happens inside a relatively long function.

I see. I usually view switch cases as similar to labels and don't mind
goto between nested switch if it jumps near the actual case of the
switch and not deep within it (it's basically like sharing the same
"case" in my mind), but point taken. I admit that mindset applies to
small switch/case blocks though, not like this function.

To be honest, I think it will be easier if I skip the no-op patch
because this will allow me to allocate the hwndText in the helper
function itself and skip passing it as argument, and encapsulate it
there along with other stuff like the bools. Of course, most of the
function will still be a cut & paste just like before (after the
realloc part) so it will only seem artificially large but not that
many changes (the control_char handling can be simplified by checking
all of them for len of zero, which is better anyway). There's no issue
with this, I hope? Or do I really have to make the no-op patch first?