Difference between revisions of "Meeting: talking debugger with Jocelyn at ETH"

(Meetings)
(Organisation)
 
Line 9: Line 9:
  
 
= Organisation =
 
= Organisation =
* To be confirmed soon
+
 
 
== Date ==
 
== Date ==
 
* arrival: tuesday 18th of December, Zurich HB either at 13h56 or 12h56
 
* arrival: tuesday 18th of December, Zurich HB either at 13h56 or 12h56
Line 22: Line 22:
  
  
References:
+
= References =
 
* http://dev.eiffel.com/CddBranch
 
* http://dev.eiffel.com/CddBranch
 
* http://se.ethz.ch/people/leitner/publications/cdd_leitner_esec_fse_2007.pdf
 
* http://se.ethz.ch/people/leitner/publications/cdd_leitner_esec_fse_2007.pdf
 
* http://portal.acm.org/citation.cfm?id=1082983.1083251
 
* http://portal.acm.org/citation.cfm?id=1082983.1083251
  
Please complete the following list with your items:
+
= Please complete the following list with your items =
  
 
* General dicussion about the CDD idea
 
* General dicussion about the CDD idea
Line 42: Line 42:
 
   (it doesn't care whether it gets the gui one or the console one).  
 
   (it doesn't care whether it gets the gui one or the console one).  
 
**** Is it okay to retrieve the whole application state through ABSTRACT_DEBUG_VALUE's? Seems to become really slow with big applications.
 
**** Is it okay to retrieve the whole application state through ABSTRACT_DEBUG_VALUE's? Seems to become really slow with big applications.
 +
Jocelyn: What do you mean by retrieving the whole application state ?
 +
however, in the debugger, there are currently too many object representations: DUMP_VALUE, DEBUGGED_OBJECT, ABSTRACT_DEBUG_VALUE ...  I feel this is too much, and I'd like to find to redesign this part. However, ABSTRACT_DEBUG_VALUE is more likely to be the more stable interface for the future. So it is best sticking with it for now.
 +
 
*** Console debugger crashes (Arno can you provide details please)
 
*** Console debugger crashes (Arno can you provide details please)
 
**** TTY debugger seems to have problems when exception in infix features occurs -> not possible to use tty menu anymore
 
**** TTY debugger seems to have problems when exception in infix features occurs -> not possible to use tty menu anymore
Line 58: Line 61:
 
**** How can we use the debugger for it  
 
**** How can we use the debugger for it  
 
***** old values
 
***** old values
 +
Jocelyn: for now, old values are not yet fully supported by the debugger, maybe for 6.3 ...
 
***** invariants
 
***** invariants
 +
Jocelyn: we hope to improve that part related to invariant soon, since for now the debugger is not helpful with invariant violation/
 +
 
**** How can we use ES capture and "Review" for it?
 
**** How can we use ES capture and "Review" for it?
 +
Jocelyn: for what ?
 +
 
** Build system needs
 
** Build system needs
 
*** How to do it properly on Windows? (Stefan, please provide details here)
 
*** How to do it properly on Windows? (Stefan, please provide details here)
Line 65: Line 73:
 
  Jocelyn: Those days, I am integrating+updating works from es-make team (Bernd and others) into the trunk,  
 
  Jocelyn: Those days, I am integrating+updating works from es-make team (Bernd and others) into the trunk,  
 
   this way it should be much more easier to build the EiffelStudio tools. Still in progress.
 
   this way it should be much more easier to build the EiffelStudio tools. Still in progress.
 +
By the way, I won't commit anything for 6.0, it would be better for you to switch to 6.1 (or trunk .. 6.2) regarding building process
 +
 
** Guided tour through debugger
 
** Guided tour through debugger
 +
Jocelyn: do you mean ... a tutorial how to use the debugger ? or you want me to present how the debugger is designed during the meeting ?
 
** Guided tour through build system
 
** Guided tour through build system
 +
Jocelyn: same question as above
 
** Unable to debug ES with ES (only Stefan and Andreas). Maybe Jocelyn can have a quick look.
 
** Unable to debug ES with ES (only Stefan and Andreas). Maybe Jocelyn can have a quick look.
 +
Jocelyn: it is definitally possible on Windows, and I used it on linux too, but maybe there are some issues with pipe sharing ... (internal to the Eiffel debugger)
 
** Capture and Replay
 
** Capture and Replay
 
*** Compare ES Capture and Replay and Sel. Capture and Replay
 
*** Compare ES Capture and Replay and Sel. Capture and Replay

Latest revision as of 03:07, 17 December 2007

This page is to plan the agenda for the meeting we plan to have with Jocelyn when he visits ETH in Zurich.

Involved parties:

  • Jocelyn
  • Bertrand
  • Andreas
  • Stefan
  • Arno

Organisation

Date

  • arrival: tuesday 18th of December, Zurich HB either at 13h56 or 12h56
  • departure: wednesday around 14h..
  • tuesday night, hotel, or hosted by ETH people (but not in the streets)

Meetings

  • tuesday: 14:00 -> 15:00 .. about build process, ... (Arno)
  • tuesday: 15:00 -> ... : debugger related points (Andreas, Arno)
  • wednesday: morning: continuation of debugger related points (Andreas, Arno starting at 10)
  • wednesday: afternoon .. Jocelyn will be on train, and ETH can do a final meeting to "compute" all the information ;)


References

Please complete the following list with your items

  • General dicussion about the CDD idea
    • Where should it go?
    • What are the biggest obstacles?
    • CDD debuggers needs
      • How to best do fg debuging of test cases
      • How to best do bg extraction of test cases
        • Is our approach of using the debugger in console mode the right one?
Jocelyn: For now, there is GUI debugger, TTY debugger, and I guess CDD debugger, 
 but you are not obliged to be in "console" mode (one can imagine a HTTP debugger, or even a silent debugger)
Andreas: The CDD test case extractor must be given an extractor 
 (it doesn't care whether it gets the gui one or the console one). 
        • Is it okay to retrieve the whole application state through ABSTRACT_DEBUG_VALUE's? Seems to become really slow with big applications.
Jocelyn: What do you mean by retrieving the whole application state ?
however, in the debugger, there are currently too many object representations: DUMP_VALUE, DEBUGGED_OBJECT, ABSTRACT_DEBUG_VALUE ...  I feel this is too much, and I'd like to find to redesign this part. However, ABSTRACT_DEBUG_VALUE is more likely to be the more stable interface for the future. So it is best sticking with it for now.
      • Console debugger crashes (Arno can you provide details please)
        • TTY debugger seems to have problems when exception in infix features occurs -> not possible to use tty menu anymore
        • Sometimes combination of other tty functions (e.g. displaying all classes in system)
Jocelyn: TTY debugger is not perfect (not enough time to work on it), what should be done is
1) change the "menu like control" by a "shell like control"
2) having the debugger controller in a separate thread, this way we could support "interrupt" and so on
3) ... many improvement, and support all the debugger functionality
  • ok, then do it :-). No, seriously, the question is if and how we can improve the tty debugger just to enable system level tests for our cdd
    • Console Menu (again Arno)
      • Multiple debuggers at the same time (large long term issue)
Jocelyn: That might be possible, but this won't be that easy for now, however, 
 I have been changing progressively the design to be more flexible, 
 and to allow that in the future. Maybe ISE could plan this for 6.3 with other major design changes.
      • Prestate capturing
        • How can we use the debugger for it
          • old values
Jocelyn: for now, old values are not yet fully supported by the debugger, maybe for 6.3 ...
          • invariants
Jocelyn: we hope to improve that part related to invariant soon, since for now the debugger is not helpful with invariant violation/
        • How can we use ES capture and "Review" for it?
Jocelyn: for what ?
    • Build system needs
      • How to do it properly on Windows? (Stefan, please provide details here)
      • (by Stefan) well, i am getting used to it. The problems with the updated build scripts is that they might not be compatible with 6.0, which cdd is based on? (i admit i did not try)
Jocelyn: Those days, I am integrating+updating works from es-make team (Bernd and others) into the trunk, 
 this way it should be much more easier to build the EiffelStudio tools. Still in progress.
By the way, I won't commit anything for 6.0, it would be better for you to switch to 6.1 (or trunk .. 6.2) regarding building process
    • Guided tour through debugger
Jocelyn: do you mean ... a tutorial how to use the debugger ? or you want me to present how the debugger is designed during the meeting ?
    • Guided tour through build system
Jocelyn: same question as above
    • Unable to debug ES with ES (only Stefan and Andreas). Maybe Jocelyn can have a quick look.
Jocelyn: it is definitally possible on Windows, and I used it on linux too, but maybe there are some issues with pipe sharing ... (internal to the Eiffel debugger)
    • Capture and Replay
      • Compare ES Capture and Replay and Sel. Capture and Replay
Jocelyn: ES provides "Execution Recording and Replay", or should I say "Review"
      • What are strengths and weaknesses?
      • Can approaches be combined for mutal benefit?
      • Where to go from here?