API review

Proposer: Rob Wheeler

Present at review:

  • List reviewers

Question / concerns / comments


  • realtime publisher/srv
    • make methods that are not part of the public api (such as "publishingLoop" etc) private.
    • document public api
    • the publishing/call loop should only lock the buffer when there is new data available. Periodically locking it causes many trylock's to fail.
  • realtime infuser
    • document public api


There's a great need of doxygen.

  • I believe that this is the only user of BufferedData* messages I would propose to pull them into this package. They are used for debugging, and as far as I can see noone is using them in code. They will be most useful for rostopic which can handle finding them.

  • The APIs need summaries of expected uses and conditions where it's appropriate to use these classes and where not.

Meeting agenda

  • talk to kevin about realtime srv
  • move recoder to .. (stu decide) ->

    • see if realtime publisher can support this functionality
    • have realtime array publisher? -> roadmap

  • realtime publisher drops msgs in buffer between realtime and non-realtime
    • can ros take msgs at 1kHz? -> done from ethercat

    • recorder limits bandwidth usage
  • infuser
    • get data from non-reltime into realtime
    • limited to 1 reader because of allocation question in rt
    • use in realtiem publisher?
    • look at rt fifo's. Josh maybe found one
    • set policy for when one side can't keep up
    • we are not using infuser: it does not support what we need for the trajectory controller
    • could create infuser that uses locks and priority inversion as we do now in trajectory controller
    • proposal:
      • write realtime box (stu)
      • kill realtime infuser


Package status change mark change manifest)

  • /!\ Action items that need to be taken.

  • {X} Major issues that need to be resolved

Wiki: realtime_tools/Reviews/2009-10-21_API_Review (last edited 2009-11-12 21:49:35 by wim)