Created attachment 53232 [details]
example

see comment in the head for how to compile

> ./states1 2>&1 --gst-debug="*SCHED*:5" | grep -v gst_pad_chain
LOG   (0x804e810 - 0:00:04.101340000)       GST_SCHEDULING(16487)
gstpad.c(810):gst_pad_set_blocked_async:<filter2:sink> blocking pad filter2:sink
LOG   (0x804e810 - 0:00:04.101489000)       GST_SCHEDULING(16487)
gstpad.c(816):gst_pad_set_blocked_async:<filter2:sink> waiting for block

means its waiting in
GST_PAD_BLOCK_WAIT (pad);

that is gstpad.c line 871 ?

if (!callback) {
  GST_CAT_LOG_OBJECT (GST_CAT_SCHEDULING, pad, "waiting for block");
  GST_PAD_BLOCK_WAIT (pad);
  GST_CAT_LOG_OBJECT (GST_CAT_SCHEDULING, pad, "blocked");
}

Created attachment 53948 [details]
updated example

* this now gets bejond the first block/unblock
* rewiring now silence the whole pipeline
* the second try to block on the pad of the newly linked element blocks again

Created attachment 54184 [details]
reduced test case

this testcase shows that it is not the segment handling of the adder that
causes the silence after rewiring pads. The new example initialy link
src1->sink then blocks on the outgoing pad of src1 and swaps src1 with src2.
After swapping it unblocks the pad and sets src to playing.

Unfortunately the output remains silent.

Created attachment 54185 [details]
reduced testcase

some cleanup, added delimiters for sections in --gst-debug output

Created attachment 54271 [details]
reduced testcase

I now run it as:
./states3 --gst-debug="basesrc:4"

this shows that initially src2 is paused as it is not linked. after swapping
src1 and src2 src1 gets paused as this one now is not linked.
I also set src2 to playing, but somehow no task gets started
(gst_base_src_start()), no gst_base_src_activate_{push,pull} call gets logged.

Created attachment 54274 [details]
reduced test case

show pad activation problem

Is this still valid with the set-flushing-unblocks-pads patch from last week?

I now use the locked state and now it works for me.