Debug/Phone 3-Dec-2009

Call Info

Meeting Title:	Conference Call on Device Debugging
Date & Time:	Thursday 3 Dec 2009 at 8:00am PDT / 10:00 am CDT / 11:00am EDT
USA/Canada tool free dial-in:	(866) 869-3090
International dial-ins:	Austria 01 92 87 547, Germany 06 92 22 21 61 10, Sweden 08 56 61 94 27, France 01 70 73 27 20, UK 02 08 11 81 026 or 08 44 57 90 592 (Complete list of international dial in numbers)
Passcode:	5107492128#

Attendees

• Wind River
  • Pawel Piech
  • Randy R.
  • Martin O.
• IBM
  • Darin
  • Jeremy
  • Naia Jane
• TI
  • Patrick
  • Navid
  • Alain
  • Dobrin
• Freescale
  • Daniel
• Nokia
  • Ken
• Ericsson
  • Marc

Agenda and Minutes

• Status
  • Pawel
    • [Bug 291267] [flex-hierarchy] Viewer does not preserve expansion state correctly when elements are inserted/removed due to IModelDelta.CONTENT
    • Breakpoints view enhancement support
  • Darin
    • Breakpoints view ehnancement
      • Reviewing functionality
      • Creating detail pane support for java breakpoints (bug Bug 290288)
        • There are difficulties in getting the user interaction correctly
        • Extended detail pane interface to have ISavable to allow detail pane to save contents before pane is closed
        • Also added a "dirty" quality, but didn't add a prompt for user
    • Deferred launch configuration template (bug Bug 41353)
      • Bug with lots of votes,
      • Came across a blocking issues
      • Ran out of budgeted time for this feature
      • Backed out changes from CVS
    • Breakpoint actions:
      • Would like to investigate to re-use platform commands instead of introducing a brand new concept of breakpoint actions.
• Breakpoint View
  • (Marc Khouzam) Flexible-hierarchy breakpoint view usage - DSF is in bug
  • Breakpoints view is (should be) functionally 100% backward compatible
  • What should be behavior on existing a "dirty" detail pane
    • Save immediately vs. drity detail pane
    • Too complicated to ask for confirmation dialog because there isn't the same infrastructure for dirty views as for dirty editors
  • Plan to commit to head after M4
• (Marc Khouzam) Bug 295658 Breakpoint actions
  • Marc: need actions on trace points. Stuck between CDT actions and non-existent platform actions
  • If it's not in M5, it will not make it into Helios
  • Marc: we have a need for action enablement expressions
  • Darin: to investigate commands in M5
  • Ken: also support using the command framework
• (Alain Lee) Bug 145635 - Pin-to-context
  • Pawel: I think we have three options to proceed here:
    1. Bury the idea and just declare that pin and clone for Debug views is just misguided.
    2. Try to resurrect the pin and clone bug in platform. This is a labor intensive and risky option. Adding a breadcrumb is one option which would address some of the concerns raised in the bug. However, using a breadcrumb would introduce its own complications in that it would need to allow debug context providers from other views besides the Debug view. Otherwise it would not be a general solution.
    3. Go for an alternative. The compare mode (http://wiki.eclipse.org/DSDP/DD/MultiContext#Proposal:_Compare_mode_in_Variables_views) is my favorite: much more elegant and than the whole pin and clone business, and it accomplish the main use case pin+clone: to compare data from different contexts. This would actually require only a small change in platform, while most of the work would be up to the debug model (i.e. DSF).
  • TI's request:
    • Alain Lee: requirement is to pin the view in one context to compare data
    • Dobrin: another use case is to track a data element of a context even when debugger needs to switch focus between other contexts.
    • GDB added reverse debugging, it may be another use case for this feature. So that we allow users to compare contexts in time.
  • Darin:
    • Debugger has a central view that controls context in all other views
    • Like the breadcrumb
    • Compare is a different workflow
    • Alternative: open a new workbench window is not feasable for users that need to track many more contexts.
    • Java debugger doesn't have many use cases to require this feature.
    • Views have global view settings. This would add complexity to the implementation.
  • Conclusion
    • Adding the breadcrumb to debugger views is the favorite idea for making the pin and clone of debugger views feature adequate for platform.
      • However there are serious technical challenges to adding the breadcrumb which means that realistically it's too late in 3.6 to try to implement this feature.
    • There is a possible simpler alternative to the proposed fix: to only override the the viewer input in the view, in order to implement the pinning of the view.
      • The limitation of this solution is that debug actions (run/step/etc) will not be updated to the view's active context as they are with the current proposed fix.
      • This alternative fix would not work in memory view.