This PR is a workaround for an issue discovered by OSS-Fuzz when trying to parse a SDP with a large number of lines.
Long story short, the g_strsplit function is not efficient when dealing with string with a lot of delimiter (\n) occurrences.
As a consequence we limit the number of SDP lines to 10000 in order to avoid the timeout on OSS-Fuzz and work around the issue in Janus.
Of course this is not the optimal solution since we do not want to hardcode a number of lines. In the long term we plan to rewrite the splitting algorithm to avoid such inefficiencies.
atoppi
commented
2 months ago
This PR is a workaround for an issue discovered by OSS-Fuzz when trying to parse a SDP with a large number of lines. Long story short, the
g_strsplitfunction is not efficient when dealing with string with a lot of delimiter (\n) occurrences.As a consequence we limit the number of SDP lines to
10000in order to avoid the timeout on OSS-Fuzz and work around the issue in Janus. Of course this is not the optimal solution since we do not want to hardcode a number of lines. In the long term we plan to rewrite the splitting algorithm to avoid such inefficiencies.