[PATCH v2] winebus.sys: Allow mapping of SDL controllers

[Aric Stewart]

With input from Zebediah Figura

Signed-off-by: Aric Stewart <aric(a)codeweavers.com>
---
 dlls/winebus.sys/bus_sdl.c | 66 +++++++++++++++++++++++++++++++++++++++++++---
 1 file changed, 63 insertions(+), 3 deletions(-)

[Zebediah Figura]

On 3/7/19 11:36 AM, Aric Stewart wrote:
> +            ULONG index = 0;
> +            KEY_VALUE_PARTIAL_INFORMATION *info;
> +            DWORD length, result_len;
> +            NTSTATUS status;
> +
> +            length = sizeof(KEY_VALUE_PARTIAL_INFORMATION) + (1000 * 
> sizeof(WCHAR));
> +            info = HeapAlloc(GetProcessHeap(), 0, length);
> +
> +            do {
> +                status = NtEnumerateValueKey(key, index, 
> KeyValuePartialInformation,
> +                                             info, length, &result_len);
> +                if (status == STATUS_SUCCESS)
> +                {
> +                    CHAR mapping[1000] = {0};
> +                    WideCharToMultiByte(CP_ACP, 0, info->Data,
> +                        (info->DataLength / sizeof(WCHAR)), mapping, 
> sizeof(mapping),
> +                        NULL, NULL);
> +                    if (mapping[0] != 0)
> +                    {
> +                        TRACE("Setting mapping 
> %s\n",debugstr_a(mapping));
> +                        pSDL_GameControllerAddMapping(mapping);
> +                    } else {
> +                        ERR("Failed to convert mapping 
> %s\n",debugstr_w(info->Data));
> +                    }
> +                    index ++;
> +                }
> +            } while (status == STATUS_SUCCESS);
> +            HeapFree(GetProcessHeap(), 0, info);
> +            NtClose(key);

I have to imagine that these strings can be arbitrarily long, and so I'd 
think it would be better to dynamically reallocate the structure.

I still don't like the use of NT APIs here: just because a real native 
driver uses it doesn't mean we have to, and they're generally more 
complex and harder to read. Still, if you insist...

[Aric Stewart]

On 3/15/19 9:47 AM, Zebediah Figura wrote:
> On 3/7/19 11:36 AM, Aric Stewart wrote:
> > +            ULONG index = 0;
> > +            KEY_VALUE_PARTIAL_INFORMATION *info;
> > +            DWORD length, result_len;
> > +            NTSTATUS status;
> > +
> > +            length = sizeof(KEY_VALUE_PARTIAL_INFORMATION) + (1000 * sizeof(WCHAR));
> > +            info = HeapAlloc(GetProcessHeap(), 0, length);
> > +
> > +            do {
> > +                status = NtEnumerateValueKey(key, index, KeyValuePartialInformation,
> > +                                             info, length, &result_len);
> > +                if (status == STATUS_SUCCESS)
> > +                {
> > +                    CHAR mapping[1000] = {0};
> > +                    WideCharToMultiByte(CP_ACP, 0, info->Data,
> > +                        (info->DataLength / sizeof(WCHAR)), mapping, sizeof(mapping),
> > +                        NULL, NULL);
> > +                    if (mapping[0] != 0)
> > +                    {
> > +                        TRACE("Setting mapping %s\n",debugstr_a(mapping));
> > +                        pSDL_GameControllerAddMapping(mapping);
> > +                    } else {
> > +                        ERR("Failed to convert mapping %s\n",debugstr_w(info->Data));
> > +                    }
> > +                    index ++;
> > +                }
> > +            } while (status == STATUS_SUCCESS);
> > +            HeapFree(GetProcessHeap(), 0, info);
> > +            NtClose(key);
>
> I have to imagine that these strings can be arbitrarily long, and so I'd think it would be better to dynamically reallocate the structure.

Agreed, patch sent.
> I still don't like the use of NT APIs here: just because a real native driver uses it doesn't mean we have to, and they're generally more complex and harder to read. Still, if you insist...

It is easier for me here as I am being passed a UNICODE_STRING with the full path for the driver registry path in an NT API format. 

-aric

[Zebediah Figura]

On 03/15/2019 02:35 PM, Aric Stewart wrote:
> It is easier for me here as I am being passed a UNICODE_STRING with the full path for the driver registry path in an NT API format.
>
> -aric

For opening the key, yes, better to use NtOpenKey(). But there's no need
to use NtEnumerateValueKey(); you can pass the same handle to
RegEnumValue(). This would also let you use RegEnumValueA() instead of
doing the conversion yourself.

[Aric Stewart]

On 3/15/19 2:45 PM, Zebediah Figura wrote:
> On 03/15/2019 02:35 PM, Aric Stewart wrote:
> > It is easier for me here as I am being passed a UNICODE_STRING with the full path for the driver registry path in an NT API format.
> >
> > -aric
>
> For opening the key, yes, better to use NtOpenKey(). But there's no need
> to use NtEnumerateValueKey(); you can pass the same handle to
> RegEnumValue(). This would also let you use RegEnumValueA() instead of
> doing the conversion yourself.

I had not thought about mixing the NT registry functions and the advapi32 ones...  That feels like crossing the streams bad. :)

-aric