[PATCH 3/4] qcap: Delay setting v4l device caps.

[Zebediah Figura]

On 8/28/20 10:52 AM, Jeff Smith wrote:
> Signed-off-by: Jeff Smith <whydoubt at gmail.com>
> ---
>  dlls/qcap/v4l.c | 20 +++++++++-----------
>  1 file changed, 9 insertions(+), 11 deletions(-)
> 
> diff --git a/dlls/qcap/v4l.c b/dlls/qcap/v4l.c
> index 6d67dd19be..c6a5e45831 100644
> --- a/dlls/qcap/v4l.c
> +++ b/dlls/qcap/v4l.c
> @@ -222,6 +222,9 @@ static HRESULT v4l_device_get_format(struct video_capture_device *iface, AM_MEDI
>  {
>      struct v4l_device *device = v4l_device(iface);
>  
> +    if (!device->current_caps)
> +        return CopyMediaType(mt, &device->caps[0].media_type);
> +
>      return CopyMediaType(mt, &device->current_caps->media_type);
>  }
>  
> @@ -406,6 +409,9 @@ static void v4l_device_init_stream(struct video_capture_device *iface)
>      ALLOCATOR_PROPERTIES req_props, ret_props;
>      HRESULT hr;
>  
> +    if (!device->current_caps)
> +        set_caps(device, &device->caps[0]);
> +

Explanation for why to do this, either as a comment in the code or as
part of a commit message, would be helpful. Patch 4/4 justifies setting
current caps to NULL until it's used, but it doesn't justify this hunk.

I'm not exactly sure we want to remove error checking here.

>      req_props.cBuffers = 3;
>      req_props.cbBuffer = device->current_caps->video_info.bmiHeader.biWidth * device->current_caps->video_info.bmiHeader.biHeight;
>      req_props.cbBuffer = (req_props.cbBuffer * device->current_caps->video_info.bmiHeader.biBitCount) / 8;
> @@ -667,14 +673,6 @@ struct video_capture_device *v4l_device_create(struct strmbase_source *pin, USHO
>      for (i = 0; i < device->caps_count; ++i)
>          device->caps[i].media_type.pbFormat = (BYTE *)&device->caps[i].video_info;
>  
> -    if (!set_caps(device, &device->caps[0]))
> -    {
> -        ERR("Failed to set pixel format: %s\n", strerror(errno));
> -        if (!have_libv4l2)
> -            ERR_(winediag)("You may need libv4l2 to use this device.\n");
> -        goto error;

The first error message is redundant, but I think the second should be
retained [and probably moved into set_caps()].

> -    }
> -
>      device->d.ops = &v4l_device_ops;
>      device->pin = pin;
>      device->state = State_Stopped;
> @@ -682,9 +680,9 @@ struct video_capture_device *v4l_device_create(struct strmbase_source *pin, USHO
>      InitializeCriticalSection(&device->state_cs);
>      device->state_cs.DebugInfo->Spare[0] = (DWORD_PTR)(__FILE__ ": v4l_device.state_cs");
>  
> -    TRACE("Format: %d bpp - %dx%d.\n", device->current_caps->video_info.bmiHeader.biBitCount,
> -            device->current_caps->video_info.bmiHeader.biWidth,
> -            device->current_caps->video_info.bmiHeader.biHeight);
> +    TRACE("Format: %d bpp - %dx%d.\n", device->caps[0].video_info.bmiHeader.biBitCount,
> +            device->caps[0].video_info.bmiHeader.biWidth,
> +            device->caps[0].video_info.bmiHeader.biHeight);

This trace isn't doing much anymore, especially since we always dump the
output of v4l_device_get_format(), so it can probably just be removed,
or replaced with a call to strmbase_dump_media_type() in
v4l_device_init_stream().

>  
>      return &device->d;
>  
> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL: <http://www.winehq.org/pipermail/wine-devel/attachments/20200828/eaabbd99/attachment.sig>