Re: Questions about CTF format

Re: Questions about CTF format

[Mathieu Desnoyers]

* Diego Dompe (ddompe@gmail.com) wrote:
> Hi,
> 
> I'm developing a custom tracer for an embedded product that will generate
> CTF format. I was able to generate generic traces that can be interpreted
> properly with babeltrace (but not with eclipse, I already file a bug for
> that), but I found the CTF specification lacking in some aspects (I had to
> peek into lttng-generated CTF traces to figure out some details). I was
> wondering what is the proper mailing list to clear my questions and provide
> feedback on the CTF specification for improvement in the areas where the
> documentation is not detailed yet. I don't see any CTF-specific mailing
> list, it's OK to discuss it here? Or maybe directly with a developer(s)?

Hi Diego,

Yes, this mailing list would be the proper place, along maybe with
adding the MCA tiwg mailing list in CC, which I'm doing here.

Thanks,

Mathieu

-- 
Mathieu Desnoyers
Operating System Efficiency R&D Consultant
EfficiOS Inc.
http://www.efficios.com

Re: Questions about CTF format

[Diego Dompe]

[-- Attachment #1.1: Type: text/plain, Size: 2042 bytes --]

Hi Mathieu,

Thanks for the help. Here is my list of details:

- Clocks: the spec doesn't explain properly that timestamps are an offset
from the base time of the clock they refer to. Since I was using 64bit
timestamps I somehow assumed that I was using absolute timestamps from the
epoch (although the spec doesn't says it either).

- I saw that the lttng-generated traces for metadata are always a multiple
of 4k in size (at least the ones I generate for either kernel or user
space). I can't find where in the spec it mentions requirements regarding
metadata packet padding. I was generating metadata packets that ended up
right after my TSDL and eclipse wasn't happy about it (although I didn't
try babeltrace). Also I found that the lttng-generated traces have a
"empty" metadata packet after the metadata containing the TSDL, I didn't
find either any documentation regarding this.

Regards,

Diego

On Thu, Jul 5, 2012 at 8:20 AM, Mathieu Desnoyers <
mathieu.desnoyers@efficios.com> wrote:

> * Diego Dompe (ddompe@gmail.com) wrote:
> > Hi,
> >
> > I'm developing a custom tracer for an embedded product that will generate
> > CTF format. I was able to generate generic traces that can be interpreted
> > properly with babeltrace (but not with eclipse, I already file a bug for
> > that), but I found the CTF specification lacking in some aspects (I had
> to
> > peek into lttng-generated CTF traces to figure out some details). I was
> > wondering what is the proper mailing list to clear my questions and
> provide
> > feedback on the CTF specification for improvement in the areas where the
> > documentation is not detailed yet. I don't see any CTF-specific mailing
> > list, it's OK to discuss it here? Or maybe directly with a developer(s)?
>
> Hi Diego,
>
> Yes, this mailing list would be the proper place, along maybe with
> adding the MCA tiwg mailing list in CC, which I'm doing here.
>
> Thanks,
>
> Mathieu
>
> --
> Mathieu Desnoyers
> Operating System Efficiency R&D Consultant
> EfficiOS Inc.
> http://www.efficios.com
>

[-- Attachment #1.2: Type: text/html, Size: 2806 bytes --]

[-- Attachment #2: Type: text/plain, Size: 155 bytes --]

_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

Re: Questions about CTF format

[Matthew Khouzam]

[-- Attachment #1.1: Type: text/plain, Size: 2634 bytes --]


On 12-07-05 10:49 AM, Diego Dompe wrote:
> Hi Mathieu,
>
> Thanks for the help. Here is my list of details:
>
> - Clocks: the spec doesn't explain properly that timestamps are an
> offset from the base time of the clock they refer to. Since I was
> using 64bit timestamps I somehow assumed that I was using absolute
> timestamps from the epoch (although the spec doesn't says it either).
This is acceptable AFAIK, just to be clear, you can use arbitrary time
origins, that means epoch is a valid origin too. You can also use a
scaling factor, just be careful applying it since a double has 53 bits
and thus you may lose precision.
>
> - I saw that the lttng-generated traces for metadata are always a
> multiple of 4k in size (at least the ones I generate for either kernel
> or user space). I can't find where in the spec it mentions
> requirements regarding metadata packet padding. I was generating
> metadata packets that ended up right after my TSDL and eclipse wasn't
> happy about it (although I didn't try babeltrace).
Could you send the trace please, I can look into it.

> Also I found that the lttng-generated traces have a "empty" metadata
> packet after the metadata containing the TSDL, I didn't find either
> any documentation regarding this.
>
> Regards,
>
> Diego
>
> On Thu, Jul 5, 2012 at 8:20 AM, Mathieu Desnoyers
> <mathieu.desnoyers@efficios.com
> <mailto:mathieu.desnoyers@efficios.com>> wrote:
>
>     * Diego Dompe (ddompe@gmail.com <mailto:ddompe@gmail.com>) wrote:
>     > Hi,
>     >
>     > I'm developing a custom tracer for an embedded product that will
>     generate
>     > CTF format. I was able to generate generic traces that can be
>     interpreted
>     > properly with babeltrace (but not with eclipse, I already file a
>     bug for
>     > that), but I found the CTF specification lacking in some aspects
>     (I had to
>     > peek into lttng-generated CTF traces to figure out some
>     details). I was
>     > wondering what is the proper mailing list to clear my questions
>     and provide
>     > feedback on the CTF specification for improvement in the areas
>     where the
>     > documentation is not detailed yet. I don't see any CTF-specific
>     mailing
>     > list, it's OK to discuss it here? Or maybe directly with a
>     developer(s)?
>
>     Hi Diego,
>
>     Yes, this mailing list would be the proper place, along maybe with
>     adding the MCA tiwg mailing list in CC, which I'm doing here.
>
>     Thanks,
>
>     Mathieu
>
>     --
>     Mathieu Desnoyers
>     Operating System Efficiency R&D Consultant
>     EfficiOS Inc.
>     http://www.efficios.com
>
>


[-- Attachment #1.2: Type: text/html, Size: 4814 bytes --]

[-- Attachment #2: Type: text/plain, Size: 155 bytes --]

_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev

Re: Questions about CTF format

[Mathieu Desnoyers]

* Diego Dompe (ddompe@gmail.com) wrote:
> Hi Mathieu,
> 
> Thanks for the help. Here is my list of details:
> 
> - Clocks: the spec doesn't explain properly that timestamps are an offset
> from the base time of the clock they refer to. Since I was using 64bit
> timestamps I somehow assumed that I was using absolute timestamps from the
> epoch (although the spec doesn't says it either).

Not sure if it was there when you wrote your comment, but the CTF spec
currently has:

"The "offset_s" and "offset" fields indicate the offset from POSIX.1
Epoch, 1970-01-01 00:00:00 +0000 (UTC), to the zero of value of the
clock."

> - I saw that the lttng-generated traces for metadata are always a multiple
> of 4k in size (at least the ones I generate for either kernel or user
> space). I can't find where in the spec it mentions requirements regarding
> metadata packet padding. I was generating metadata packets that ended up
> right after my TSDL and eclipse wasn't happy about it (although I didn't
> try babeltrace). Also I found that the lttng-generated traces have a
> "empty" metadata packet after the metadata containing the TSDL, I didn't
> find either any documentation regarding this.

Babeltrace had a limitation on not accepting packets smaller than 4kB
(page size actually), but it is now fixed.

Thanks,

Mathieu

> 
> Regards,
> 
> Diego
> 
> On Thu, Jul 5, 2012 at 8:20 AM, Mathieu Desnoyers <
> mathieu.desnoyers@efficios.com> wrote:
> 
> > * Diego Dompe (ddompe@gmail.com) wrote:
> > > Hi,
> > >
> > > I'm developing a custom tracer for an embedded product that will generate
> > > CTF format. I was able to generate generic traces that can be interpreted
> > > properly with babeltrace (but not with eclipse, I already file a bug for
> > > that), but I found the CTF specification lacking in some aspects (I had
> > to
> > > peek into lttng-generated CTF traces to figure out some details). I was
> > > wondering what is the proper mailing list to clear my questions and
> > provide
> > > feedback on the CTF specification for improvement in the areas where the
> > > documentation is not detailed yet. I don't see any CTF-specific mailing
> > > list, it's OK to discuss it here? Or maybe directly with a developer(s)?
> >
> > Hi Diego,
> >
> > Yes, this mailing list would be the proper place, along maybe with
> > adding the MCA tiwg mailing list in CC, which I'm doing here.
> >
> > Thanks,
> >
> > Mathieu
> >
> > --
> > Mathieu Desnoyers
> > Operating System Efficiency R&D Consultant
> > EfficiOS Inc.
> > http://www.efficios.com
> >

-- 
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com

Questions about CTF format

[Diego Dompe]

[-- Attachment #1.1: Type: text/plain, Size: 684 bytes --]

Hi,

I'm developing a custom tracer for an embedded product that will generate
CTF format. I was able to generate generic traces that can be interpreted
properly with babeltrace (but not with eclipse, I already file a bug for
that), but I found the CTF specification lacking in some aspects (I had to
peek into lttng-generated CTF traces to figure out some details). I was
wondering what is the proper mailing list to clear my questions and provide
feedback on the CTF specification for improvement in the areas where the
documentation is not detailed yet. I don't see any CTF-specific mailing
list, it's OK to discuss it here? Or maybe directly with a developer(s)?

Regards,

Diego

[-- Attachment #1.2: Type: text/html, Size: 771 bytes --]

[-- Attachment #2: Type: text/plain, Size: 155 bytes --]

_______________________________________________
lttng-dev mailing list
lttng-dev@lists.lttng.org
http://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev