[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [hyades-dev] Committer vote required for M10 checkin



Eugene,  I agree with these changes.

Richard K. Duggan
Problem Determination Enablement
IBM Toronto Laboratory
External: 905-413-2396
Internal: 969-2396



                                                                           
             Eugene                                                        
             Chan/Toronto/IBM@                                             
             IBMCA                                                      To 
             Sent by:                  hyades-dev@xxxxxxxxxxx              
             hyades-dev-admin@                                          cc 
             eclipse.org                                                   
                                                                   Subject 
                                       [hyades-dev] Committer vote         
             05/25/2004 05:05          required for M10 checkin            
             PM                                                            
                                                                           
                                                                           
             Please respond to                                             
                hyades-dev                                                 
                                                                           
                                                                           





The following M10 defects require Committer's approval: (I vote YES)

Defect 63508:
bugzilla_63508 -- Console view has wrong combination of buttons enabled

Rationale:

There used to be a bug on the execution side preventing the terminated
signal from reaching the UI. This has since been fixed; the only remaining
thing is to send the platform a signal saying the IProcess has terminated.
The console and debug views will update themselves accordingly.

Risk Assessment:

No API change. The fix is to send the DebugEvent.TERMINATE signal on our
IProcess when we get the processExited() event from our internal process.
Low risk change.


Defect 63484:
bugzilla_63484 -- Profiling types not being notified of launch in some
cases

Rationale:

This fix is to add notification for the profiling types during launch for
those launch delegates that were not doing so. This signal is vital for
some profiling types in order to perform pre-launch setup. This was missed
during implementation of some of our profiling launchers (e.g. Java Applet,
Attach - Java Process).

Risk Assessment:

No API change. The fix is to query the selected profiling set and notify
each selected type before performing the launch.
Low risk change.


Defect 63289:
bugzilla_63289 -- Hosts disappear from the host list in the Profiling
Config window after adding new hosts

Rationale:

This fix updates the host list in preferences when the user adds a host
from the launch configuration dialog tab. This was missed during the move
to the new launch system.

Risk Assessment:

No API change. The fix is to update the preferences whenever the user adds
a host in the hosts tab.
Low risk change.


Defect 63272:
bugzilla_63272 -- Profiling tab appears where it shouldn't

Rationale:

This fix applies to the run/debug mode launch shortcuts in the Profiling
and Logging perspective (only from the toolbar launch icon). Because of a
platform deficiency, new perspectives cannot inherit existing launch
shortcuts; we have to define them again. These re-definitions were pointing
to the profiling implementation of the launch shortcuts instead of the
run/debug implementations, so it behaved as though the user launched via
the profile launch shortcut.

Risk Assessment:

No API change. Simple fix: switch to the existing run/debug implementation
in plugin manifest.
Low risk change.


Defect 62274:
bugzilla_62274 -- Profiling type attributes should not be stored in launch
configuration

Rationale:

This is a fix for the way profiling sets were incorrectly stored. While
most of the contents of the set (e.g. which types are selected) are stored
in preferences where they should be, the configuration attributes from the
profiling types were stored in the launch configuration. Because of this,
any changes to a profiling type that involved a change in attributes
affected only one launch configuration. The change should affect all launch
configurations that use that set.

Risk Assessment:

No API change. The fix is to store attributes in preferences with the rest
of the set definition, and to no longer store them in launch
configurations. Fix verified by the probekit team.
Low/Medium risk change.


Defect 59965:
bugzilla_59965 -- Show instance level information not taken into account

Rationale:

This is a fix for the specific scenario where the user selects only the
Execution Time Analysis profing type and selects "Show instance level
information". In this case, the launcher was sending the wrong options to
the rac: TRACE_MODE=full and ALLOCATION_INFORMATION=none. In this case, we
do want allocation information.

Risk Assessment:

No API change. The fix is to upgrade the value of ALLOCATION_INFORMATION to
normal when the user checks the "Show instance level information" box.
Low risk change.

Defect 59371:
bugzilla_59371 -- Profile launch UI sometimes has a blank line in
Profiling...

Rationale:

This fixes the problem of the blank line showing up in the profiling sets
list, and the strange selection behavior (click on one item, another is
selected). Although no functionality is missing, this is very annoying to
the user.

Risk Assessment:

No API change. The blank line was caused by circumventing the viewer in
some cases and modifying the table widget directly, which confuses the
viewer managing the table (viewers are not meant to be used this way). The
fix is to do all the work on the table through the viewer. The strange
selection behavior was caused by querying the table widget's selection
index (same problem). The items were automatically sorted by the viewer,
but not in the original list of sets, so we were looking up the wrong
index.

Low/Medium risk change.


Thanks for your time.
--------------------------------------------------------------------
Eugene Chan
IBM Toronto Laboratory
D3/ENW/8200/MKM
Voice:   1-905-413-6102
Email:  ewchan@xxxxxxxxxx