On Tuesday 28 February 2017 02:46 AM, Steven Rostedt wrote:

On Mon, 27 Feb 2017 21:52:11 +0530 Pratyush Anand panand@redhat.com wrote:

...

Hi Steve,

On Tuesday 21 February 2017 12:17 AM, Steven Rostedt wrote:

...

...

Can you please help me with some pointer so that I can parse them correctly.

Take a look at the latest kbuffer-parse.c and kbuffer.h. That's the code that does the parsing. You should be able to use that code directly and simply just use those functions instead of having to parse it yourself.

Thanks for the pointer. It was helpful.

I have added an option in trace-cmd itself which can help to detect *approximate* peak memory.Not sure on which mailing list I can send the RFC for trace-cmd. BTW my rough patch is here. Before it can be polished, an early feedback would be helpful.

https://github.com/pratyushanand/trace-cmd/commit/d3710da68f8eecf4d99a1c49c5...

Feel free to just post it ;-) I think having code polished in the open is always the best.

Yes,yes..I wasn't just sure that which mailing list should I use. OK, I will use same mailing list and cc list as of this patch.

Anyway, the few things that I will have you change is the changing of static to global. There may be functions that are already global that you can use. Or at least, we should change the names. The

OK.

"host_is_big_endian" can be derived from the pevent. I'm assuming that you'll require creating a pevent.

HUmm..tracecmd_alloc_fd() populates pevent->host_bigendian. However, this code does not work on a trace data acquired in a file, rather it works on live streamed trace buffer. So tracecmd_alloc_fd() will not be called. However, I can use tracecmd_host_bigendian() from trace-cmd.h.

Anyway, let me post the patch and we can discuss it over there.

Thanks for your time.

~Pratyush