Re: [cdt-debug-dev] MI Future work

[Elena Zannoni <ezannoni@xxxxxxxxxx>]

Alain Magloire writes:

Hi Alain,

 > 
 > Tom @ redhat com started a few words on this, but since the agenda
 > for the meeting was full, the subject was not expanded.
 > 
 > The state of things.  CDT 1.0.x will use gdb 5.2.1 or any
 > debugger providing MI level 1 as define by gdb 8-).
 > The debugger(gdb) is actually started with the options "-mi1"
 > to make sure it will fail upfront, for example it will not work
 > reliably with gdb-5.0(mi0) and break in a lot of places.
 > 
 > However, for the future we would like :
 > 
 > - to take advantages of the new "mi2" but still work with "mi1".
 >   To do this, the code need some refactoring to isolate common parts
 >   from dependent mi level.
 > 

We have dropped mi0 from the FSF sources, but mi1 and mi2 are there.

 > - on  the gdb side, it would be __really__ nice if the we could
 >   query gdb about the current default level of mi it supports:
 > 
 >   (gdb) -mi-current-level
 >   ^done,value=1
 >   (gdb)
 > 
 >   or something.  This could allow the CDT/MI plugin to query
 >   about the level and adjust the parser at runtime, say we
 >   have parser for mi 1 and mi 2.
 > 

true, when I get the interprete stuff integrated, I can go
back and add an -info-interpreter command, or similar. 

 > - gdb-5.4 (mi 2) has better oob support to provide feedback.
 >   For example, one problem with the current mi, it does not provid
 >   enough information for the event thread handler to make a decision.
 > 
 > - keith @ redhat was working on some interpreter, we would like
 >   to see to take advantage to finish the prompt interface.
 > 

I have started integrating Keith's and Apple's work into the current
FSF gdb tree. It will be in before we branch for 5.4.

Thanks for the update.

Elena


 > 
 > We've received a __lot__ of request/features/enhancement from
 > our customers, QNX internal tools group, Eclipse CDT/Bugzilla.  It
 > goes from having things like:
 > 
 > ..  Something similar to DDD, to visualize variables
 > 
 > to
 > 
 > ... debugging a process that is fork()ing whether to take
 >     advantage of gdb catchpoints or some other mechanism
 > 
 > and the holy grail
 > 
 > ... debugging JNI code with the C debugger and the Java debugger
 >     working together. 
 > 
 > 
 > We are reviewing the feedbacks and looking at what's feasible
 > for 1.1.x(or what ever we call it).  To late for 1.0.x.
 > If more ideas, please forward to the list or bugzilla or  etc ....
 > 
 > 
 > 
 > BTW, many thanks to redhat for always reacting to our complaints
 > about gdb/mi and fixing bugs.  Thanks Alifiya, for some good Q/A 8-)
 > 
 > 
 > _______________________________________________
 > cdt-debug-dev mailing list
 > cdt-debug-dev@xxxxxxxxxxx
 > http://dev.eclipse.org/mailman/listinfo/cdt-debug-dev