[Theora-dev] Question about MVect in decode.c

[Alen Ladavac]

Apparently, this is by design. The RTC fails when the value is copied to
another variable. Explanation from original authors (IINM) is that in the
case where this variable is not initialized, the target variable is not
used, later on. So this is not a real error. On the other hand, clearing the
variable prior to usage is supposed to have performance impact (not sure
that someone actually bothered to measure it, though). So they decided to
leave it that way. You should disable RTC only on that library, or if you
are statically linking to Theora, then you might want to clear that
variable, just in debug version (RTC is not enabled in release versions).

So, answers to your questions would be:
1. No, it will not lead to faulty operation.
2. It is not fixed and is not likely to be fixed. The behavior is by design.

HTH,
Alen

----- Original Message -----
From: "LS" <linesoft at plotinka.ru>
To: <theora-dev at xiph.org>
Sent: Monday, May 30, 2005 11:16 AM
Subject: [Theora-dev] Question about MVect in decode.c


> Good day!
>
> Could you please explain me the following:
> In case of playback of video stream theora, MS Visual Studio .NET 2003
shows a warning message in every frame about the
> usage of not initialized variable MVect in the file decode.c (run-time
> check).
> In case of key-frames playback no warning appears.
> When I disable the run-time check, the video stream plays normally, but it
is undesirable. Theora is implemented
> in my project as a source, so when I disable the run-time check it will
disable it for the while project. I would
> like to avoid it if possible.
>
> Questions:
> 1. can this variable (if used before being initialized) lead to the
cosec's faulty operation?
> 2. Is this bug fixed in the latest releases? If fixed, where can I get the
correct source files?
>
> Thanks,
> LS
>
> _______________________________________________
> Theora-dev mailing list
> Theora-dev at xiph.org
> http://lists.xiph.org/mailman/listinfo/theora-dev
>