Re: [diamon-discuss] [lttng-dev] CTF2-PROP-1.0: Proposal for a major revision of the Common Trace Format, version 1.8

[Philippe Proulx <pproulx@efficios.com>]

----- Original Message -----
> From: "Jérémie Galarneau" <jeremie.galarneau@efficios.com>
> To: "Philippe Proulx" <pproulx@efficios.com>
> Cc: "diamon-discuss" <diamon-discuss@lists.linuxfoundation.org>, "lttng-dev" <lttng-dev@lists.lttng.org>,
> "tracecompass-dev" <tracecompass-dev@eclipse.org>, "etienne bergeron" <etienne.bergeron@gmail.com>, "francois doray"
> <francois.doray@gmail.com>
> Sent: Sunday, 11 December, 2016 06:13:09
> Subject: Re: [lttng-dev] CTF2-PROP-1.0: Proposal for a major revision of the Common Trace Format, version 1.8
>
> May I propose the addition of packet-begin and packet-end attributes,
> analogous to the current "timestamp_begin" and "timestamp_end" fields?
> 
> These fields greatly facilitate the job of readers that need to index
> and seek within a CTF stream (time-wise) if clocks are used. Moreover,
> the attributes could be re-used shall the index format currently
> produced by LTTng become part of the official specification in one
> form or another.
> 
> Thoughts?

Isn't it enough to tag the `timestamp_begin` field with
`update-data-stream-clock-now` and the `timestamp_end` field with
`update-data-stream-clock-after-packet`?

If you need to seek, read the packet header and context, then
look at the current values of the data stream clocks and at their
projected values (after packet).

Philippe Proulx
EfficiOS Inc.
http://www.efficios.com/

> Jérémie