[PATCH] gdi32: Allow any mode passed to SetBkMode().

[Jeff Smith]

On Wed, Oct 7, 2020 at 10:53 AM Nikolay Sivov <bunglehead at gmail.com> wrote:
>
>
>
> On Wed, 7 Oct 2020 at 18:35, Jeff Smith <whydoubt at gmail.com> wrote:
>>
>> On Wed, Oct 7, 2020 at 2:24 AM Huw Davies <huw at codeweavers.com> wrote:
>> >
>> > On Tue, Oct 06, 2020 at 11:22:29PM -0500, Jeff Smith wrote:
>> > > Signed-off-by: Jeff Smith <whydoubt at gmail.com>
>> > > ---
>> > > Testing revealed that SetBkMode() does not treat 'invalid' values as one
>> > > might assume.  Discovered this when working on another issue.  Note that
>> > > fixing this to match Windows simplifies one of the tests for that bug.
>> >
>> > Which then begs the question: How should these invalid modes be
>> > treated visually?
>>
>> FWIW, testing In Windows points to the default being transparent.
>> That implies that any wine code that checks this value should check
>> for equality/non-equality with OPAQUE to be "safe".
>
>
> What ends up in metafile record for invalid modes?

It contains the unaltered value for both valid and invalid modes, as
a little-endian 32-bit signed int of course.

Tested with -1 0 1 2 3 0xdeadbeef.

I tested on wine, with the patch applied, and got matching results.


>> > Since there doesn't appear to be a real application that
>> > depends on this, we should keep things as they are.
>>
>> Would be nice (and removes 8 unnecessary lines of code), but no
>> real harm in leaving it as it is.
>>
>> Thanks,
>> Jeff
>>