This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC 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: [PATCH] Re: scheduling vs lexical scopes


> On Sun, Dec 30, 2001 at 03:37:09PM +0100, Jan Hubicka wrote:
> > > On Sun, Dec 30, 2001 at 12:51:54PM +0100, Jan Hubicka wrote:
> > > > > > (but it will cause many fragments to be created)
> > > > > 
> > > > > I do worry about this, particularly for non-dwarf3 debug targets.
> > > > > Not sure what to do about it though.
> > > > I was thinking about prunning the regions a bit - recognizing instrucitons
> > > > not interesting for symbolic level debugger (those that do not trap nor
> > > > access memory nor are just after line number note) and do not create
> > > > fragments for these. I guess on riscy machines this should help a bit
> > > > as typical reg-reg operation fits this category.
> > > 
> > > Is reading memory interesting to symbolic level debugger?
> > 
> > I guess it can have watchpoint on it assuming that the memory is something
> > accessible from symbolic level (de-facto all memory?)
> > 
> > > On the other side, changing a /v pseudo definitely is interesting.
> > 
> > Can symbolic debugger stop somehow on such instruction? Can I have
> > watchpoint on register?
> 
> Sure, e.g. using next or step. I doubt you can have watchpoint on a
Step will bring me to the next line number note.
I am not sure if we need to have everything consistent in the next - stepping
trought uncompletted statement looks like something where I would not
require all variables to make 100% sense.

Honza
> register, but that doesn't matter.
> 
> 	Jakub


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