Attention is currently required from: pespin.

falconia has posted comments on this change. ( https://gerrit.osmocom.org/c/osmo-bts/+/32988 )

Change subject: refactor: replace rtppayload_is_valid() with preening before enqueue ......................................................................

Patch Set 1:

(2 comments)

File src/common/l1sap.c:

https://gerrit.osmocom.org/c/osmo-bts/+/32988/comment/f0de429a_646eb8d4 PS1, Line 1237: default:

you are dropping the AMR path here. Why? […]

I am not removing the logic of suppressing AMR bwe, I simply moved it to the rtp_payload_input_preen() function that executes before enqueueing, rather than after dequeueing.

File src/common/osmux.c:

https://gerrit.osmocom.org/c/osmo-bts/+/32988/comment/d978b405_9b5eee00 PS1, Line 413: /* We have to apply the same checks as in l1sap_rtp_rx_cb(), in case

I understood this extra byte was only appearing in HR?

Yes, as well as an extended FR/EFR format of my own invention (falconia/rtp_traulike branch) which I will be addressing much later, certainly after discussing it in OsmoDevCall, not before.

Osmux is AMR always,

I did not know this detail! It was certainly one of those "how am I supposed to know this arcane detail" moments...

does it really apply here?

The header octet stripping logic is never needed for AMR - so if you say that Osmux is AMR only, then I need to rethink and redesign this patch.