I don't know enough about RTCP to review this... but if it talks about rounding, always going to the next 32bit word seems less likely than the ceiling round that is currently done ;) But then, I have no idea about that code :)

The code is parsing a list, at the end of the list there is a "0". When the "0" is reached the offset should be changed to be the first byte after the current 32 bit word.
So if the current offset is 4 then that is pointing to the first byte in the second 32 bit word and so after this operation offset should be 8 which is the first byte in the third 32bit word.
(4+3)&~3 = 4  (wrong)
(4+4)&~3 = 8  (write)

Created attachment 185101 [details] [review]
corrected the author

Ok, thanks for the explanation. Makes sense :)

commit 1ad98b0d98c8701273b5e6e4e42b9daff4f40bec
Author: Pascal Buhler <pascal.buhler@tandberg.com>
Date:   Tue Aug 24 13:14:33 2010 +0200

    rtcpbuffer: Round to next 32bit word, not current 32bit word at end of SDES