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: stack-protector guard location




Sent from my iPhone

On Aug 27, 2009, at 6:48 PM, Roland McGrath <roland@redhat.com> wrote:

Well, given enough debugger hacking, I suppose that could be
made to work in this instance at the moment.  But it would
probably break when we have need of another anonymous variable
type of thing for some other feature.

Granted. My initial thought was that we might add some sort of custom
attribute to mark this. But that's likely to be even further from Just Works.


I guess I just assumed that the debugger really wouldn't get
too confused because the variable name isn't likely to overlap
with anything that's currently in use.  I do like the current
patch because your use cases Just Work with the current gdb.

I actually didn't mean "confused gdb" so much as "make gdb confuse its
user". i.e. "info locals" showing __stack_guard__ is a dubious feature.
Frankly "p &__stack_guard__" is a rather dubious to me. Nothing (what
nothing? well, hardly anything) says my program can't have a real variable
called __stack_guard__.
Except that name is implemention namespace which is reserved.



Anyway, it doesn't have a name. That's just lying. It ain't right.


Perhaps no DW_AT_name but DW_AT_description of "stack-guard".

But hey, an opinion from the debugger camp is certainly welcome.

I'm not a debugger camper but I impersonate one in compiler camps.



Thanks, Roland


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