Re: [PATCH v5 10/30] tools/lib/traceevent: Man page for tep_register_trace_clock()

[Steven Rostedt <rostedt@goodmis.org>]

On Fri, 12 Apr 2019 16:37:51 +0300
Tzvetomir Stoyanov <tstoyanov@vmware.com> wrote:

> Create man page for tep_register_trace_clock() libtraceevent API.
> 
> Signed-off-by: Tzvetomir Stoyanov <tstoyanov@vmware.com>
> ---
>  .../Documentation/libtraceevent-reg_clock.txt | 77 +++++++++++++++++++
>  1 file changed, 77 insertions(+)
>  create mode 100644 tools/lib/traceevent/Documentation/libtraceevent-reg_clock.txt
> 
> diff --git a/tools/lib/traceevent/Documentation/libtraceevent-reg_clock.txt b/tools/lib/traceevent/Documentation/libtraceevent-reg_clock.txt
> new file mode 100644
> index 000000000000..8bf6c5348fcf
> --- /dev/null
> +++ b/tools/lib/traceevent/Documentation/libtraceevent-reg_clock.txt
> @@ -0,0 +1,77 @@
> +libtraceevent(3)
> +================
> +
> +NAME
> +----
> +tep_register_trace_clock - Registers the clock, used for the timestamp of
> +the events.
> +
> +SYNOPSIS
> +--------
> +[verse]
> +--
> +*#include <event-parse.h>*
> +
> +int *tep_register_trace_clock*(struct tep_handle pass:[*]_tep_, const char pass:[*]_trace_clock_);
> +--
> +
> +DESCRIPTION
> +-----------
> +The _tep_register_trace_clock()_ function registers the clock, used for
> +the timestamps of the events. When an event is recorded, a timestamp is attached
> +to it. This timestamp comes from a specified clock. Supported clocks can be seen
> +in tracefs/trace_clock, the selected one is in square brackets.
> +The _tep_ argument is the trace event parser context. The _trace_clock_ argument
> +is the name of the clock, the string is copied internally.
> +
> +RETURN VALUE
> +------------
> +The _tep_register_trace_clock()_ function returns 0 in case of success.
> +If there is not enough memory to copy the _trace_clock_ string -1 is returned
> +and errno is set to ENOMEM.
>

Hmm, this is more code I think we should remove from libtraceevent, and
have the tools just set the tep flag to print in usecs or not. The
library shouldn't be setting policy.

The tools (perf, trace-cmd, etc) should just implement their own
tep_register_trace_clock() (not calling it that), and set the USEC flag
depending on what it wants.

-- Steve