Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cdt-debug-dev] Register reading in debugger

> 
> > 
> > This is a multipart message in MIME format.
> > --=_alternative 0035E2C4C3256C91_=
> > Content-Type: text/plain; charset="us-ascii"
> > 
> >       Hi all!
> > 
> > I'm using CDT Debugger and when I want to look at all registers I've 
> > noticed that Debugger asks for value of one register two times. I think 
> > this doesn't accelerate Debugger's work and so I have small question which 
> > can be considered as suggestion of improving Debugger's work . 
> > 
> >   May be it would be better to read  all registers by GDB/MI command 
> > "-data-list-register-values x" and then parse output inside CDT Debugger. 
> > May be I'm wrong but I think that  this way would be work faster than many 
> > calls of external program(i.e. gdb). 
> > 
> 
> The code does not maintain a cache of the result, so what happen is that
> the CDT/Debug framework is also probing a second time for possible 
> children, getChildren().
> 
> For example on x86 xmm? registers are represented as structure(it shows a +
> in the Variable view).
> 
> Before registers were maintain by the Var Objects gdb mechanism, gdb was
> maintaining it for us and notify on every updates etc ..
> Unfortunately using registers has var objects in gdb was making gdb very
> unstable.  So we fall back to -data-list-register-xx.
> 
> It certainly can be made smarter for things like getChildren() or getTypeName()
> which will never change and cache this the first time we get a response. 
> 

To make things clearer, you should probably PR this in bugzilla as a bug
or an enhancement.  I will gladly take a patch if you have one.
Not sure if this will be fix in the 1.0.1 branch,
is this some sort of performance bottleneck for you ?



Back to the top