* Yao Qi ([email protected]) wrote: > On 02/27/2013 08:23 PM, Mathieu Desnoyers wrote: >> I'm curious to see what you are trying to achieve by grouping events >> this way ? > > We propose to GDB community to save GDB trace data into CTF and read CTF > data saved by GDB (via libbabeltrace). > > [PATCH 0/5, 2nd try] CTF Support > http://sourceware.org/ml/gdb-patches/2013-02/msg00669.html > > The GDB trace data is composed by several trace frames, and each trace > frame is composed by several blocks. When saving GDB trace data into > CTF, we save the block as ctf event, and save these events belong to the > same trace frame into one packet, so that the information of the trace > frame can be saved into packet context. > > When reading CTF data, usually we will look for events in current trace > frame (packet), so that I am asking this question, to see if it is > easier to do iteration in one packet.
You might want to consider an alternative design: consider this mapping instead: one event = one trace frame the event "payload" would be a sequence of blocks. Each block can contain its own information. You could use the "variant" type if you need different types for each block within a trace frame. This would fit more nicely with CTF. Thanks, Mathieu > >> >> You know you could simply put each logical event group into a different >> "channel" within the same trace, right ? > > Can you elaborate a little please? I am quite new to CTF and babeltrace. > >> >> The level of abstraction at which the babeltrace iterator presents the >> trace does not expose details about packet switching. I'd need more >> information to understand the value of your use-case. > > Hope my description above is clear. > > -- > Yao (齐尧) -- Mathieu Desnoyers EfficiOS Inc. http://www.efficios.com _______________________________________________ lttng-dev mailing list [email protected] http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
