[PATCH 0/1] MR6547: uxtheme: Better detect the presence of default transparent colour in prepare_alpha().

[Paul Gofman (＠gofman)]

Fixes a regression introduced by
c0bf67636ab3de46c67b8d02e44fb39b5853384a.

[PATCH 1/1] uxtheme: Better detect the presence of default transparent colour in prepare_alpha().

[Paul Gofman]

From: Paul Gofman

Fixes a regression introduced by
c0bf67636ab3de46c67b8d02e44fb39b5853384a.
---
 dlls/uxtheme/msstyles.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/dlls/uxtheme/msstyles.c b/dlls/uxtheme/msstyles.c
index aa78c372d15..1efd78f2295 100644
--- a/dlls/uxtheme/msstyles.c
+++ b/dlls/uxtheme/msstyles.c
@@ -1218,12 +1218,37 @@ static BOOL prepare_alpha (HBITMAP bmp, BOOL* hasAlpha, BOOL *hasDefaultTranspar
     }

if (dib.dsBm.bmBitsPixel != 32)
+    {
+        if (dib.dsBm.bmBitsPixel <= 8)
+        {
+            RGBQUAD p[256];
+            HDC hdc = CreateCompatibleDC(NULL);
+            HBITMAP prev = SelectObject(hdc, bmp);
+            UINT count = GetDIBColorTable(hdc, 0, 256, p);
+
+            SelectObject(hdc, prev);
+            DeleteDC(hdc);
+
+            for (n = 0; n < count; ++n)
+            {
+                if (RGB(p[n].rgbRed, p[n].rgbGreen, p[n].rgbBlue) == DEFAULT_TRANSPARENT_COLOR)
+                {
+                    *hasDefaultTransparentColour = TRUE;
+                    return TRUE;
+                }
+            }
+        }
         return TRUE;
+    }

/* If all alpha values are 0xff, don't use alpha blending */
     for (n = 0, p = dib.dsBm.bmBits; n < dib.dsBmih.biWidth * dib.dsBmih.biHeight; n++, p += 4)
+    {
         if ((*hasAlpha = (p[3] != 0xff)))
             break;
+        if (RGB(p[0], p[1], p[2]) == DEFAULT_TRANSPARENT_COLOR)
+            *hasDefaultTransparentColour = TRUE;
+    }

if (!*hasAlpha)
         return TRUE;

[Paul Gofman (＠gofman)]

Fixes a regression described here: https://gitlab.winehq.org/wine/wine/-/merge_requests/6401#note_82961 (where the theme has 8 bit paletted bitmap with default transparent colour which was displayed as pink instead of being transparent after the blamed commit).

[Zhiyi Zhang (＠zhiyi)]

Zhiyi Zhang (@zhiyi) commented about dlls/uxtheme/msstyles.c:
>  }
>
>  if (dib.dsBm.bmBitsPixel != 32)
> - {
> -
>    if (dib.dsBm.bmBitsPixel <= 8)

I think you need to check 16-bit bitmaps as well.

[Zhiyi Zhang (＠zhiyi)]

Zhiyi Zhang (@zhiyi) commented about dlls/uxtheme/msstyles.c:
>  if (!bmp || GetObjectW( bmp, sizeof(dib), &dib ) != sizeof(dib))
>      return FALSE;
>
>  if (dib.dsBmih.biCompression != BI_RGB)

The regression probably applies to other bitmap compression formats as well. For example, BI_BITFIELDS. But then the other formats seem unused by theme files so they might be safe to ignore.

[Paul Gofman (＠gofman)]

On Tue Sep 24 00:44:46 2024 +0000, Zhiyi Zhang wrote:
> The regression probably applies to other bitmap compression formats as
> well. For example, BI_BITFIELDS. But then the other formats seem unused
> by theme files so they might be safe to ignore.

I guess BI_BITFIELDS should coincide with the 16-bit case you mentioned?

[Zhiyi Zhang (＠zhiyi)]

On Tue Sep 24 00:55:32 2024 +0000, Paul Gofman wrote:
> I guess BI_BITFIELDS should coincide with the 16-bit case you mentioned?

I think BI_RGB for 16-bit is also valid. Please see https://learn.microsoft.com/en-us/windows/win32/api/wingdi/ns-wingdi-bitmapi....