Re: [FFmpeg-devel] [PATCH] avcodec/decode: use avcodec_get_hw_config()

From: Kacper Michajlow <kasper93@gmail.com>
To: "FFmpeg development discussions and patches"
	<ffmpeg-devel@ffmpeg.org>,
	"Kacper Michajłow" <kasper93@gmail.com>
Subject: Re: [FFmpeg-devel] [PATCH] avcodec/decode: use avcodec_get_hw_config()
Date: Mon, 8 Jul 2024 16:42:42 +0200
Message-ID: <CABPLASTAeojGt=464GQ9kCvRJpx8rnQSZPYvqgoiKad3sRrWDg@mail.gmail.com> (raw)
In-Reply-To: <171982896778.21847.1498044631421448952@lain.khirnov.net>

On Mon, 1 Jul 2024 at 12:16, Anton Khirnov <anton@khirnov.net> wrote:
>
> Quoting Kacper Michajłow (2024-06-02 03:38:18)
> > Fixes libavcodec/decode.c:1035:61: runtime error: member access within
> > null pointer of type 'const struct AVCodecHWConfigInternal'.
> >
> > This can happen when hwaccel fails to initialize and hw_configs[i] is
> > NULL.
>
> How? All this should be runtime-constant.

As we can read in hw_configs description.

> Array of pointers to hardware configurations supported by the codec, or NULL if no hardware supported.
> The array is terminated by a NULL pointer.

It uses for (i = 0;; i++), so it always gets the `public` address.
Since public is the first member of `AVCodecHWConfigInternal`, it
silently works; the offset of `public` is zero, so it remains NULL,
and the check in the next line rejects it. Nevertheless, it is better
to check it correctly as avcodec_get_hw_config() does.

The "fails to initialize" part in the commit message is misleading. I
think I just triggered this part of the code on some fallback path,
but the value of hw_configs[i] is not affected by this.

- Kacper
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".