Re: Question on LTTng status...

* Michael Boerner ([email protected]) wrote:
> I hate to ask this question but I have to, it seems like that the activity
> has been decreasing since 2005, is this just a misconception?  Finally, I am
> trying to confirm whether some changes have been made like whether you have
> resolved the format incompatibility between revisions or do does it still
> exist?
> 

Hello,

Thanks for asking this. In fact, I have been quite busy working on the
project, but activity on ltt-dev did not reflect it.

Format incompatibility between revisions are to be expected, since this
is still under heavy development. I document the compatible versions on
the website (http://ltt.polymtl.ca > LTTng+LTTV versions compatibility).
This is the tradeoff between letting the development process go on and
letting users find easily what versions will work together.

The real way to get something stable and well integrated with the Linux
kernel will be to get merged into the Linux kernel mainline tree. This
is precisely what I am working on. You can have a look at my posts on
LKML to figure out where most of my time is spent. ;) So rather than
adding new features, I am currently polishing LTTng, the markers and
low-level programming primitives such as local_cmpxchg in order to get
my work integrated into mainline. Note that the cannot happen in only a
few weeks, and what is expected is that I generalize every pieces of
infrastructure that I have done for my custom tracing needs so that
other users in the kernel tree can use them.

Also, I am currently dropping the "facilities" for a more generic
per-marker selection of events. There will still be "groups", but it
won't have such a big impact on how probes are connected to modules. One
of the major changes is to trade the 8 bit facility id, 8 bit event id
for a dynamically allocated 16 bits event id (marker id could be a
synonym). It will be much more flexible, allowing to enable/disable each
marker individually at runtime, and select the channel they must be
written to. This will, indeed, break compatibility with older versions
since this is a change to the trace format.

Also, I switched from git to quilt to manage my work, which may make it
rather more difficult to follow my WIP on the website. The nice side of
it is that it is much easier for me to send it upstream (to Andrew
Morton and on LKML for review) because it is splitted in nice logical
patches easier to review.

Regards,

Mathieu

> MEB
> 
> _______________________________________________
> Ltt-dev mailing list
> [email protected]
> http://listserv.shafik.org/mailman/listinfo/ltt-dev
> 

-- 
Mathieu Desnoyers
Computer Engineering Ph.D. Student, Ecole Polytechnique de Montreal
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68