[lttng-dev] lttng questions

[Philippe Proulx]

On Wed, Jul 19, 2017 at 3:05 PM, Nalin Pai <nalinpai at gmail.com> wrote:
> More questions.
>
> I notice that filter expression parser does not validate to check if the
> field-name in the expression is valid or whether the expression value is
> consistent with the field type (e.g check that expression value is integer
> if field type is integer). Are there any plans of stricter checking of
> expressions? If the user makes a mistake while typing out the expression is
> there any feedback to indicate the error?

It cannot be indicated to the user at event rule creation time
(lttng enable-event) because future events could match an event rule
you create now.

When a field's type is not what it should be according to the event
rule's filter expression, then the whole filter expression evaluates to
0 (false), thus the whole event rule is considered not satisfied.

>
> Thanks,
> Nalin
>
> On Wed, Jul 19, 2017 at 10:11 AM, Nalin Pai <nalinpai at gmail.com> wrote:
>>
>> Thanks for your answers. I have a question on event filters. We will be
>> representing IPv6 addresses as arrays of bytes. We would like to be able to
>> filter on IPv6 address fields. Does lttng filtering on arrays and sequences?
>> If yes, then what's the filter expression format for this.

This is an upcoming feature, probably part of 2.11, but I could be wrong
(about the version).

Phil

>>
>> Thanks,
>> Nalin
>>
>>
>>
>> On Sun, Jul 16, 2017 at 4:49 PM, Philippe Proulx <eeppeliteloop at gmail.com>
>> wrote:
>>>
>>> On Sun, Jul 16, 2017 at 2:28 PM, Nalin Pai <nalinpai at gmail.com> wrote:
>>> >
>>> > Hello,
>>> >
>>> > We are looking at using lttng as a solution for all our trace and debug
>>> > requirements. We want to get away from traditional file based string format
>>> > logging and move to lttng, which then gives us the flexibility of dumping
>>> > the traces to file or the network. With this background, I have several
>>> > questions for the team.
>>> >
>>> > 1) CTF supports struct fields. I don’t see any ctf_ macros in lttng
>>> > that will allow us to define tracepoints having struct fields. Is there a
>>> > way to do this that I am missing?
>>>
>>> Complex CTF types (except a few very specific exceptions) are not
>>> supported in LTTng-UST, the user space LTTng tracer. If you need a
>>> structure data type, then you need to "explode" it, perhaps with field
>>> name prefixes, as of the latest version of LTTng. For example:
>>> `my_struct_user_id`, `my_struct_msg_id`, `my_struct_msg`.
>>>
>>> >
>>> > 2) Display hints. Sometimes there is a need to display certain fields
>>> > in a particular way, e.g. IP and MAC addresses. While the tracepoint
>>> > definition for these fields are going to be integers or sequences of
>>> > integers, we would like babeltrace to display these fields in a certain
>>> > format. Short of modifying babeltrace to recognize these fields by their
>>> > names and handle them specially, is there any other way of achieving the
>>> > same?
>>>
>>> In the short- to medium-term, with Babeltrace 2.0 (currently in "pre"
>>> development stage), you can create your own sink component to do
>>> something with events. The default textual display is called
>>> sink.text.pretty. If you need a modified version of this sink
>>> component class, you could copy its source and create your own
>>> modified plugin which, as you wrote, would recognize the fields by
>>> name, etc. You could then run the CLI as such:
>>>
>>>     babeltrace /path/to/trace
>>> --component=sink.my-plugin.my-component-class
>>>
>>> I can't comment on Trace Compass/LTTng Scope.
>>>
>>> I see no other way for the moment. CTF 2 will have a way to indicate
>>> that specific fields are to be displayed using a given format string,
>>> for example, but its specification is not available currently (only a
>>> proposal is available here
>>> <http://diamon.org/ctf/files/CTF2-PROP-1.0.html>). Babeltrace 2.1 will
>>> support CTF 2, after which a future version of LTTng will produce it.
>>>
>>> Phil
>>>
>>> >
>>> > -Nalin
>>> >
>>> >
>>> >
>>> >
>>> > _______________________________________________
>>> > lttng-dev mailing list
>>> > lttng-dev at lists.lttng.org
>>> > https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
>>> >
>>
>>
>