[lttng-dev] some events lost

[Sébastien Barthélémy]

Hello Mathieu,

2011/12/30 Mathieu Desnoyers <compudj at krystal.dyndns.org>:
>> From another log system I can tell the traced program is going well,
>> so this is probably a problem with the way I use LTTng.
>> [...]
>> The target system is running a 2.6.33.9 kernel, so the futex workaround
>> is active. Could it be the cause of this gap?
>
> I doubt it.

ok

>> Or maybe the buffers get full? But I have not seen any message telling
>> so. I tried the
>> --subbuf-size (I'm not sure what a subbuffer is though) with no
>> success: lttng list reported no
>> change in the subbuffer size.

I was trying to change the subbuffer size while the channel was in use.
It now makes sense.

I just did a few test with increased buffer size and have seen no more gap.

> However, the information is available in the trace files, so I have a TODO item
> somewhere on my todo list to make babeltrace print a warning when it
> figures out that events have been dropped.

That would be a good think, we need to know if we can trust reconstructed states
or not.

Thank you once again

Best regards

-- Sébastien