[gst-devel] Scheduler requirements

[Benjamin Otte]

On Sun, 4 Apr 2004, Andy Wingo wrote:

> Hm, it seems that's the way to go. Personally I don't think malloc
> should be allowed at all within processing functions.
>
The way I'd want to go is to introduce a flag that elements may set when
they don't use malloc and obey realtime characteristics (say
GST_ELEMENT_REALTIME). So you could build your app by just using
elements that set this flag. This would of course require marking
GStreamer functions as ok/not ok for realtime usage in the docs.

We will definitely not get around using malloc in elements, because we use
external libs and those often don't care about realtime at all.

Benjamin