Re: [Savonet-users] crash tests, how to fix the 4th? FEEDBACK

[David B. <dav...@gm...>]

Hi,

I'm back from holidays, but I'll be gone tomorrow again. In the
meantime, here's a little note about your experiments.

> Main playlist uses mksafe [...]
> If main fails, moves to rescue but only play silence. Dont know why

If you wrap the main playlist with mksafe, you get an infallible
source. Indeed, mksafe() plays silence instead of failing. This is
probably not wanted: it's better to let the main playlist fail, and
have your fallback() handle the failure in a smart way.

This is what you do in these two cases:

> Main playlist is fallible
> Rescue playlist is mksafe = creates the mount point, moves to the rescue playlist when main is deleted empty or corrupted. ALL OK

> Main playlist is fallible
> Rescue playlist uses playlist.safe = creates the mount point, moves to the rescue playlist when main is deleted empty or corrupted. ALL OK

In both cases you need an infallible rescue source. If you get it
using playlist.safe() that's nice, but liquidsoap refuses to start if
the rescue playlist is empty or invalid. If you get it by using
mksafe() around a normal playlist, that works too, but it means that
you'll get silence when the rescue fails. In this last solution I'm
not sure what the interest is compared to simply having a single
playlist wrapped in a mksafe().

Have fun.
-- 
David