[PATCH v2 0/1] MR6249: dinput: Use the correct array index in keyboard_create_device().

[Aida Jonikienė (＠DodoGTA)]

This fixes a segfault when launching NFS Underground.

Fixes: f434ea12b83 ("dinput: Implement DIPROP_SCANCODE.")

--
  v2: dinput: Use the correct array index in keyboard_create_device().

https://gitlab.winehq.org/wine/wine/-/merge_requests/6249

[PATCH v2 1/1] dinput: Use the correct array index in keyboard_create_device().

[Aida Jonikienė]

From: Aida Jonikienė

This fixes a segfault when launching NFS Underground.

Fixes: f434ea12b83 ("dinput: Implement DIPROP_SCANCODE.")
---
 dlls/dinput/keyboard.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/dlls/dinput/keyboard.c b/dlls/dinput/keyboard.c
index 8ec9dd47150..7d063308d69 100644
--- a/dlls/dinput/keyboard.c
+++ b/dlls/dinput/keyboard.c
@@ -211,7 +211,7 @@ HRESULT keyboard_create_device( struct dinput *dinput, const GUID *guid, IDirect

if (FAILED(hr = dinput_device_init_device_format( &impl->base.IDirectInputDevice8W_iface ))) goto failed;

-    for (i = 0, index = 0; i < 512; ++i)
+    for (i = 0, index = 0; i < impl->base.device_format.dwNumObjs; ++i)
     {
         if (!GetKeyNameTextW( i << 16, instance.tszName, ARRAY_SIZE(instance.tszName) )) continue;
         if (!(dik = map_dik_code( i, 0, subtype, impl->base.dinput->dwVersion ))) continue;

[Aida Jonikienė (＠DodoGTA)]

> If there's a regression it's more likely related to the addition of DIPROP_SCANCODE support.

I reverted all of the `DIPROP_SCANCODE` switch case changes and the segfault still happens (so the loop is definitely introducing/exposing the problem)

[Aida Jonikienė (＠DodoGTA)]

After some testing I discovered the segfault is caused by the 0001 patch of dinput-scancode staging patchset (and the 0002 patchset introduces some test failures)

Enabling scancode auto-detection solves both of these issues though

I have a different version of this patch that doesn't regress mainline Wine (but it still causes test failures/tests succeeding inside of a todo block on staging) :frog:

[Rémi Bernon (＠rbernon)]

So there is no upstream regression then?

[Aida Jonikienė (＠DodoGTA)]

On Mon Aug 12 06:13:44 2024 +0000, Rémi Bernon wrote:
> So there is no upstream regression then?

I guess upstream is okay-ish (but there might still be potential OOB array access for object_properties)

[PATCH v2 0/1] MR6249: dinput: Use the correct array index in keyboard_create_device(). - closed

[Rémi Bernon (＠rbernon)]

This merge request was closed by Rémi Bernon.