[Live-devel] Tricky bug in RTSP-on-HTTP server implementation?

Ross Finlayson finlayson at live555.com
Sat Oct 23 15:17:09 PDT 2010


>the new version works correctly (no more ignored RTSP requests), but 
>it seems to introduce a new bug, this time in memory management. 
>Glibc seems to crash on a double free or corruption when the RTSP 
>connection is abruptedly closed (I could see this behaviour on VLC 
>everytime I closed the client clicking on the "close" button on the 
>VLC window - I didn't have had time to analyze the network dump yet, 
>and I don't know if VLC sends a TEARDOWN or closes the socket 
>directly in that case).

It sends a "TEARDOWN", I think.  You should be able to get the same 
effect with "openRTSP" with the "-d <duration>" option.


>Please note that this happens *not only* on HTTP-over-RTSP 
>connection, but on RTSP/UDP and RTSP/TCP sessions too.

In that case, the bug could not have been introduced in the new 
version (because the changes in that version affected only 
RTSP-over-HTTP streams).

How repeatable is this??  Can you repeat this with 
"live555MediaServer" (or "testOnDemandRTSPServer") with unmodified 
source code?  Using "openRTSP -d <duration>" as a client?  (Or just 
with VLC.)  And does the crash happen every time?

If the crash happens only with modified server code, then there's not 
much I'll be able to do at my end to track it down.
-- 

Ross Finlayson
Live Networks, Inc.
http://www.live555.com/


More information about the live-devel mailing list