Antoher way of dealing with EOS. This proposal does not use the recursion to propagate the EOS signal. This implies t...

Original commit message from CVS:
Antoher way of dealing with EOS. This proposal does not use the recursion
to propagate the EOS signal. This implies that an element cannot deny an
EOS signal anymore but since the signal is generated when a NULL buffer is
pushed, somebody did something wrong anyway.
This commit is contained in:
Wim Taymans 2001-01-13 19:04:18 +00:00
parent f18c4056f7
commit 69b40b46d2

73
docs/random/wtay/eos5 Normal file
View file

@ -0,0 +1,73 @@
case 1)
(--------------------------------------------------)
! bin !
! (--------) (--------) (--------) !
! ! fakesrc! !identity! !fakesink! !
! ! 1 src ----- sink 1 src ---- sink 1 ! !
! (--------) (--------) (--------) !
! !
! (--------) (--------) (--------) !
! !fakesrc ! !identity! !fakesink! !
! ! 2 src ----- sink 2 src ---- sink 2 ! !
! (--------) (--------) (--------) !
(--------------------------------------------------)
just two scheduled chains in a single bin.
.scheduling.
we have two chains here fs1-i1-fsk1 and fs2-i2-fsk2
fakesrc1 will output 1 buffer, fakesrc2 3 buffers
.eos.
fs1 will return a NULL buffer on EOS. the scheduler
detects the NULL buffer and puts chain1 in EOS so that it
doesn't get scheduled again. The scheduler fires the EOS
signal for all the elements in the chain and puts the fs1
element in the READY state.
the same procedure happens for chain2. At that point, the bin
has no chains to schedule and so it fires EOS.
case 2)
(---------------------------------------------------)
! bin !
! (--------) (--------) (---------) !
! ! disksrc! ! mpg123 ! !audiosink! !
! ! 1 src ----- sink 1 src ---- sink 1 ! !
! (--------) (--------) (---------) !
! !
! (--------) (--------) (---------) !
! !disksrc ! ! mpg123 ! !audiosink! !
! ! 2 src ----- sink 2 src ---- sink 2 ! !
! (--------) (--------) (---------) !
(---------------------------------------------------)
same as case 1, an mp3 mixer where the mixing is performed
in hardware (in the audiosink).
we want to show how the EOS state can be undone.
.scheduling.
we have two chains here ds1-mpg1-ask1 and ds2-mpg2-ask2
the song in chain1 is about to end, chain2 has just begun
playing a song.
.eos.
fs1 will return a NULL buffer on EOS. the scheduler
detects the NULL buffer and puts chain1 in EOS so that it
doesn't get scheduled again. The scheduler fires the EOS
signal for all the elements in the chain and puts the fs1
element in the READY state.
the app sets the new location for the disksrc. The element is
put in the PLAYING state, the chain is marked !EOS so that
it will be scheduled again.