This is the mail archive of the gdb-patches@sources.redhat.com mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [rfc] Annotation level THREE


> > > Why do you need display-{begin,end}? -display-insert has been made > > > redundant by the varobj stuff - it lets the GUI efficiently track its > > > display values outside of the CLI. > > > > Variable objects don't auto-display. You seem to have to type
> > -var-evaluate-expression each time the program stops. > > Why do you need auto-display? What are you using it for?
> > Remember, every time the target changes (e.g., from a user modifying a > variable or register), the display needs to be updated. This is because > that variable/register has the potential to modify every single value > being displayed. Further, unless your using some sort of changes-only > mechanism, such as provided by the varobj, the display windows are just > not going to scale.


You're right. If the user assigns a value to a variable, using my code, the
display window for that variable won't update in Emacs until the next
statement is executed.

Even simple things, like trying to compute ``1 + 2'' (rather than ``1+2'') won't work without MI.


What do you mean by `not going to scale'?

There are a numer of factors that determine IDE performance (as defined by single-step):


- ide screen update (gui overhead)
- ide <-> gdb overhead
- gdb <-> inferior
- system load/overhead

If the IDE refreshes all display elements (instead of those that have changed) then single-step performance will be dominated by the number of visual elements that need to be refreshed. Any ide<->gdb overhead will be in the noise.

> Currently if I display an array
slice, say just a few elements from a large array, I need to parse them from
output for the whole array. Could I arrange for GDB to just output the elements
I want using variable objects?

I don't know. Perhaphs ask this on gdb@ as a specific question.


Can I get GDB to tell me what the current list of variable objects is?

The assumption is that the IDE is tracking the mapping between each of its local display elements and the corresponding varobj.


> Can
I generate names for them automatically?

I'm not sure what you mean. Looking at the doco


-var-create - * i

automatically assigns the varobj name.

Andrew



> > > The testsuite is a good source of > > > varobj examples (unfortunatly lacking from the doco):
> > > http://sources.redhat.com/gdb/current/onlinedocs/gdb_25.html#SEC565
> > > > The lack of documentation for (and apparent completeness of) GDB/MI is part of
> > the problem.
> > Um, the varobj is documented. It just lacks a vew concrete examples. > Those can be found by examining the testsuite. Both Apple and Eclipse > are using this part of the MI.


Its a question of resources. I'm one person doing things in my spare time.
Apple and Eclipse have full time teams dedicated to the task.

> > > The above list also contains thing like field-{begin,end}, > > > array-section-{begin,end} et.al. Why are they needed.
> > > > I use them to parse the output. They could probably go, if necessary, but
> > others that you plan to take out *are* needed.
> > Can you please be more specific?


Perhaps I could turn that question round. Which annotations are you planning
to keep?

Nick




Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]