Redundant Logging on User Extensions

[Chris Johns]

On 30/10/2014 8:06 am, Joel Sherrill wrote:
>
> On 10/29/2014 4:01 PM, Gedare Bloom wrote:
>> On Wed, Oct 29, 2014 at 2:42 PM, Joel Sherrill
>> <joel.sherrill at oarcorp.com> wrote:
>>> Hi
>>>
>>> As Jennifer has reviewed the output from an SMP capture
>>> test, it is clear that something which was no big deal in
>>> the old fixed record uniprocessor version is now an issue.
>>> Most extensions have an "actor" and "acted upon" thread.
>>> For example, when a task is created, the calling task and
>>> new task are important. The actor and acted upon threads
>>> were logged in separate records.
>>>
>>> In an SMP system sorted by time, these two records can
>>> become disconnected. Since we now have variable length
>>> records, it makes sense to consolidate these into a single
>>> capture record.
>>>
>>> Any complaints?
>>>
>> You probably want to support multiple ways to view the capture log,
>> including sorting by core, or by some notion of "system call" e.g. the
>> entire chain of events for a single thread creation.
> Agreed but in this case there are two records used to represent one logical
> event like "task A created task B" or "task A restarted task B". We just
> want to combine two records into one in the log.

This is fine. I suspect it will not be the last time something like this 
comes up.

> How it is sorted for display purposes is a user interface issue.

Or how an import format supports it [1].

Jennifer, any interesting traces to tease us with ?

Chris

[1] http://www.efficios.com/ctf