Attention is currently required from: laforge, pespin.
manawyrm has posted comments on this change by manawyrm. ( https://gerrit.osmocom.org/c/osmo-e1d/+/43643?usp=email )
Change subject: octoi: rework frame_rifo logic to decompress frames on the output side ......................................................................
Patch Set 5:
(3 comments)
Commit Message:
https://gerrit.osmocom.org/c/osmo-e1d/+/43643/comment/dcb1ee3e_0f674ee2?usp=... : PS4, Line 9: The previous code tried to decompress frames against the last received frame regardless of the ordering (even when the frames got re-ordered by the network) and also when a frame didn't even arrive yet (rendering the jitter buffer almost useless).
cosmetic: the commitlog lines should be wrapped at (ideally) 72 characters
Acknowledged
File src/octoi/e1oip.c:
https://gerrit.osmocom.org/c/osmo-e1d/+/43643/comment/69c13c6a_a06b9c75?usp=... : PS5, Line 263: memcpy(frame_buf, iline->e1t.last_frame, BYTES_PER_FRAME);
any reason to do the zeroing further up now?
Not quite sure what you're asking. The logic for this has moved over to frame_rifo_out() in frame_rifo.c line 164.
Or are you asking about the stack buffer allocation with the "= { 0 };"? That's leftover from an old experiment -- as long as the ts_mask is set correctly and the logic is correct, it shouldn't matter at all.
File src/octoi/octoi.c:
https://gerrit.osmocom.org/c/osmo-e1d/+/43643/comment/9b60e3e2_7fac145d?usp=... : PS5, Line 140: /* substitute with last received frame */
No substitution now afaict?
That's what the new memcpy() in frame_rifo_out() line 152 is for. Maybe the comment should be changed to "substituted"?