Skip to main content

Notice: This Wiki is now read only and edits are no longer possible. Please see: https://gitlab.eclipse.org/eclipsefdn/helpdesk/-/wikis/Wiki-shutdown-plan for the plan.

Jump to: navigation, search

Difference between revisions of "Trace Compass/Project Meetings/2016-04-01"

Line 1: Line 1:
== Agenda ==
+
== Attendees ==
 +
 
 +
* Alexandre Montplaisir (EfficiOS)
 +
* Jonathan Rajotte (EfficiOS)
 +
* Matthew Khouzam (Ericsson)
 +
* Patrick Tassé (Ericsson)
 +
* Marc-André Laperle (Ericsson)
 +
* Jean-Christian Kouamé (Ericsson)
 +
* Bernd Hufmann (Ericsson)
 +
* Geneviève Bastien (Polytechnique)
 +
 
 +
== Minutes ==
  
 
=== Go over AIs of last meeting ===
 
=== Go over AIs of last meeting ===
  
* EclipseCon writeup
+
* Bernd made a post about important Neon dates.
* ML post about important Neon dates
+
** https://dev.eclipse.org/mhonarc/lists/tracecompass-dev/msg00532.html
 +
* Marc-André hasn't published the EclispeCon writeup yet, will do so soon.
  
 
=== Efficios's planned tasks for Neon ===
 
=== Efficios's planned tasks for Neon ===
  
 +
Alex is planning to finish up and push to Gerrit the following two branches:
 
* UST 2.8 Symbol provider (for source lookup and callstack view)
 
* UST 2.8 Symbol provider (for source lookup and callstack view)
 
* LTTng-Analyses integration
 
* LTTng-Analyses integration
 +
 +
ideally before mid-April.
 +
 +
=== Callstack View and Symbol Provider additions ===
 +
 +
While working on the new UST symbol provider, Alex hit some limitations of the current symbol lookup implementation:
 +
* There is no notion of timestamps: libraries can get loaded and unloaded some one address may not point to the same symbol for the whole duration of the trace.
 +
* There is no notion of process: a trace may contain events from more than one process/executable. The symbol lookup, even the addresses themselves, are completely different from one process to another.
 +
 +
The aforementioned branch will include some patches to add the notion of process to the callstack analysis and view, as well as new methods in the symbol provider that take a "address, timestamp, pid" tuple.
  
 
=== LTTng Control MI XSD validation ===
 
=== LTTng Control MI XSD validation ===
  
* XSD validation is more a "unit test" for LTTng itself. Viewers shouldn't have to do this.
+
* According to LTTng devs, XSD validation is more a "unit test" for LTTng itself. Viewers shouldn't have to do this.
 +
* It was decided to drop the XSD validation in Trace Compass.
 +
** We won't need to import / do a CQ for the 2.8 XSD
 +
* However it would be useful for the MI protocol to have its own version, so that we can know if/when it does break. Jonathan mentioned it should be very easy to add.
 +
** https://bugs.lttng.org/issues/1008
  
 
=== Update on LTTng Live Reading ===
 
=== Update on LTTng Live Reading ===
  
* https://bugs.eclipse.org/bugs/show_bug.cgi?id=486728
+
* Due to several limitations, it was confirmed to disable the LTTng live tracing feature for the 2.0 release. See the discussion in:
 +
** https://bugs.eclipse.org/bugs/show_bug.cgi?id=486728
 +
* Bernd also pointed out that it might be good to get feedback on actual use cases before going further ahead with this.
 +
 
 +
=== New LTTng Common plugin ===
 +
 
 +
* Geneviève prepared a patch to introduce a new lttng2.common plugin.
 +
* This plugin would contain things common to both UST and kernel traces, like some aspect definitions.
 +
* Everyone was fine with it.
 +
 
 +
== Action Items ==
 +
 
 +
* None

Revision as of 18:40, 1 April 2016

Attendees

  • Alexandre Montplaisir (EfficiOS)
  • Jonathan Rajotte (EfficiOS)
  • Matthew Khouzam (Ericsson)
  • Patrick Tassé (Ericsson)
  • Marc-André Laperle (Ericsson)
  • Jean-Christian Kouamé (Ericsson)
  • Bernd Hufmann (Ericsson)
  • Geneviève Bastien (Polytechnique)

Minutes

Go over AIs of last meeting

Efficios's planned tasks for Neon

Alex is planning to finish up and push to Gerrit the following two branches:

  • UST 2.8 Symbol provider (for source lookup and callstack view)
  • LTTng-Analyses integration

ideally before mid-April.

Callstack View and Symbol Provider additions

While working on the new UST symbol provider, Alex hit some limitations of the current symbol lookup implementation:

  • There is no notion of timestamps: libraries can get loaded and unloaded some one address may not point to the same symbol for the whole duration of the trace.
  • There is no notion of process: a trace may contain events from more than one process/executable. The symbol lookup, even the addresses themselves, are completely different from one process to another.

The aforementioned branch will include some patches to add the notion of process to the callstack analysis and view, as well as new methods in the symbol provider that take a "address, timestamp, pid" tuple.

LTTng Control MI XSD validation

  • According to LTTng devs, XSD validation is more a "unit test" for LTTng itself. Viewers shouldn't have to do this.
  • It was decided to drop the XSD validation in Trace Compass.
    • We won't need to import / do a CQ for the 2.8 XSD
  • However it would be useful for the MI protocol to have its own version, so that we can know if/when it does break. Jonathan mentioned it should be very easy to add.

Update on LTTng Live Reading

  • Due to several limitations, it was confirmed to disable the LTTng live tracing feature for the 2.0 release. See the discussion in:
  • Bernd also pointed out that it might be good to get feedback on actual use cases before going further ahead with this.

New LTTng Common plugin

  • Geneviève prepared a patch to introduce a new lttng2.common plugin.
  • This plugin would contain things common to both UST and kernel traces, like some aspect definitions.
  • Everyone was fine with it.

Action Items

  • None

Back to the top