Re: MI to access value of $ prefixed variable name

On Tue, 25 May 2021 at 10:04, Simon Marchi <[email protected]> wrote:

> On 2021-05-25 12:35 a.m., Jonah Graham wrote:
> > Hello,
> >
> > I have a user of Eclipse CDT who is having problems displaying a C
> variable
> > called $j (dollar sign prefixed variable) -
> > https://bugs.eclipse.org/bugs/show_bug.cgi?id=573703
> >
> > I understand this is because $ prefixed variables are treated specially
> by
> > GDB.
> >
> > 47-stack-list-locals --thread 1 --frame 0 1
> > 47^done,locals=[{name="$f",value="0"}]
> >
> > 52-var-create --thread 1 --frame 0 - * $f
> > 52^done,name="var3",numchild="0",value="void",type="void",has_more="0"
> >
> > Note type "void" above, $f is an int in the C code, but is indeed void
> (as
> > expected) in GDB. -stack-list-locals shows the correct value, but I can't
> > use the returned name back to -var-create because of the naming conflict.
> >
> > There wasn't anything I saw in the help (
> >
> https://sourceware.org/gdb/onlinedocs/gdb/GDB_002fMI-Variable-Objects.html
> )
> > on this topic.
>
> Huh, I didn't know this was possible:
>
> https://gcc.gnu.org/onlinedocs/gcc/Dollar-Signs.html#Dollar-Signs


The main place I have seen this in practice is generated code.


> I didn't look in depth, but I don't think there's a way to make it work.
> I don't remember seeing a way to escape that dollar sign.
>

No problem - thanks for the answer.

I'd say that the simplest solution would be: don't use dollar sign in
> identifiers.  But if somebody really wants GDB to support it, then they
> can always send a patch for it (I don't think this is a priority for
> anyone).
>

OK. I will update the bugzilla in CDT with this information.

Jonah