Trace Compass/Project Meetings/2016-04-29

Attendees

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

Minutes

Go over AIs of last meeting

• Alex was quite busy with the ongoing LTTng-Analyses work, didn't have time to review the serialization patches. Will do so soon.
  • Important first item is the custom state values
  • Branch has topic custom_sv on Gerrit

• Alex and Loïc held a meeting to discuss state system API changes:
  • New algorithms by Loïc for range and "two-dimensional" queries require the logic to be in the StateSystem object directly instead of the utils method (not a wrapper around single queries anymore).
  • Alex suggested to add the new APIs to ITmfStateSystem, but keep the old methods in Utils, perhaps @Deprecated, so it doesn't cause an API break. Those could be removed eventually.
  • No rush on this part, so it will be done after Neon.

• Loïc posted his summary of state system changes to the mailing list:
  • https://dev.eclipse.org/mhonarc/lists/tracecompass-dev/msg00577.html

• Alex verified if the source code lookup action did look for files in the workspace
  • The fall-back in the code (in TmfEventsTable) is still there, so it should fall-back to looking for the file name into the workspace.
  • Marc-André remembers testing it at some point
  • Bernd will try it again to make sure it is still working. If it is not, it would be a bug.

Status of Ongoing Development for Neon

Ericsson

• Integration of XML data-driven analyses
  • Feature merged, JC is working on tests.

• Some views are not populated when they are part of an experiment
  • Work ongoing
  • Won't all be ready for Neon. For Neon, in cases of multiple traces of the same type in an experiment, only one of them will be shown.
  • Maybe have a way to manually populate a view from a single trace that is part of an experiment

• Update LTTng Control to 2.8 with current existing features
  • Done, current features are all supported
  • Includes some MI changes

Poly

All wanted patches are currently up on Gerrit.

• XML API to be made all internal
  • A big patch, but it's mostly just renames.
• API for parallelize analysis bug (isQueriable)
• Custom state values
• IO analysis and view
• Cleanup analysis requirement API

EfficiOS

• LTTng 2.8 Debug Info analysis
  • A new test trace, and updated tests, will be required.
  • New patches will be sent next week to update for the new event names and new PIC/non-PIC distinction.

• LTTng-Analyses integration (no API changes)
  • Lots of work was done by the team at EfficiOS over the previous weeks.
  • Still a few things to merge in the working branch, should be finished today (Friday).
  • Alex will update the branch on Gerrit with the final, ready-for-review, version early next week.

Organization of Project View

Should the new element be called "Views", and where should it be in the list.

• Everyone is receptive to potentially moving the "Views" element to the top of the list.
• This is very easy to change, so it should not block the current patches, but should be decided before the final release!
• Geneviève suggested running it through some users at the Poly meeting next week.

Open bugs related to new PID 0 modelling

• Bug 491276 - Analysis: AttributeNotFoundException in LttngWorker
• Bug 491348 - Analysis: Incorrect CPU tool tip in Control Flow view for swapper process
• Bug 492404 - VM analysis generates NumberFormatException on thread 0

Patches are up for the bottom two.

Geneviève mentioned she has a patch for the Critical Path analysis which may happen to fix the first one.

Problem is under control overall.

Neon Release Preparation

After RC1, the Ericsson team will run through their standard manual tests procedure.

• Many new features, will require many new tests
• User documentation
• Test spec is available on Google Docs: 2.0.0-TraceCompassTestCases

Action Items

• Alex to really review custom state value patches
• Alex + Matthew to review and merge state system node cache patches
  • Makes the cache static, reduces memory usage a lot, would be great to have for Neon.
  • Two versions exists, we can merge the two patches one after another, for easier testing/reverting if needed.

• Polytechnique meeting next week (Thursday/Friday). Most people should be there.