Checking variable scope

Checking variable scope

[Rob Quill]

Hi,

I am trying to make it possible to be able to check if a variable is
in scope at a given time. My current implementation for this involves
adding a parameter to parse_expression, and the parsing functions it
calls, all the way until it gets to the line:

if (sym == 0)
    error ("No symbol \"%s\" in specified context.", copy_name ($3));

in c-exp.y

as which point it doesn't call error, but calls another function which
returns a value signifing that the variable is not in scope.

Whilst this is OK, in that it (should) work, I was wondering if anyone
had a neater way of implementing it?

Thanks for your time.

Rob

Re: Checking variable scope

[Jim Blandy]

"Rob Quill" <rob.quill@gmail.com> writes:
> I am trying to make it possible to be able to check if a variable is
> in scope at a given time. My current implementation for this involves
> adding a parameter to parse_expression, and the parsing functions it
> calls, all the way until it gets to the line:
>
> if (sym == 0)
>    error ("No symbol \"%s\" in specified context.", copy_name ($3));
>
> in c-exp.y
>
> as which point it doesn't call error, but calls another function which
> returns a value signifing that the variable is not in scope.
>
> Whilst this is OK, in that it (should) work, I was wondering if anyone
> had a neater way of implementing it?

Why not call lookup_symbol directly?  You can get the block that needs
with block_for_pc_sect.  Are you selecting the location by source
position, or PC, or what?

Re: Checking variable scope

[Rob Quill]

Hi,

I am stepping through the program, so I suppose I'd be looking to see
if the variable is in scope at the location of the PC.

How would I call lookup_symbol, given that information? I had a brief
look, but I wasn't sure what parameters to pass to the function, or
how to get the right values.

Thanks for your help.

Rob

On 22/11/06, Jim Blandy <jimb@codesourcery.com> wrote:
>
> "Rob Quill" <rob.quill@gmail.com> writes:
> > I am trying to make it possible to be able to check if a variable is
> > in scope at a given time. My current implementation for this involves
> > adding a parameter to parse_expression, and the parsing functions it
> > calls, all the way until it gets to the line:
> >
> > if (sym == 0)
> >    error ("No symbol \"%s\" in specified context.", copy_name ($3));
> >
> > in c-exp.y
> >
> > as which point it doesn't call error, but calls another function which
> > returns a value signifing that the variable is not in scope.
> >
> > Whilst this is OK, in that it (should) work, I was wondering if anyone
> > had a neater way of implementing it?
>
> Why not call lookup_symbol directly?  You can get the block that needs
> with block_for_pc_sect.  Are you selecting the location by source
> position, or PC, or what?
>

Re: Checking variable scope

[Jim Blandy]

"Rob Quill" <rob.quill@gmail.com> writes:
> I am stepping through the program, so I suppose I'd be looking to see
> if the variable is in scope at the location of the PC.
>
> How would I call lookup_symbol, given that information? I had a brief
> look, but I wasn't sure what parameters to pass to the function, or
> how to get the right values.

You can pass NULL for 'symtab' and 'is_a_field_of_this', and
VAR_DOMAIN for domain.  block is the scoping block containing the PC.
I don't know where you're calling things from, so I can't really tell
you how to get the current PC.

Re: Checking variable scope

[Rob Quill]

On 27/11/06, Jim Blandy <jimb@codesourcery.com> wrote:
>
> "Rob Quill" <rob.quill@gmail.com> writes:
> > I am stepping through the program, so I suppose I'd be looking to see
> > if the variable is in scope at the location of the PC.
> >
> > How would I call lookup_symbol, given that information? I had a brief
> > look, but I wasn't sure what parameters to pass to the function, or
> > how to get the right values.
>
> You can pass NULL for 'symtab' and 'is_a_field_of_this', and
> VAR_DOMAIN for domain.  block is the scoping block containing the PC.
> I don't know where you're calling things from, so I can't really tell
> you how to get the current PC.
>

The aim would be to add a command to gdb, in_scope (or similar) which
can be used from a GDB script to check if a variable is in scope
without throwing an error if is isn't. As currently if I do "print a"
and a is not in scope then I get an error and the script stops, which
is fine. But even better would be if I were able to check if a was in
scope and only print (or do other things with it) if it was.

Does that help explain where I would be calling things from? As I
understand it, current when you do "print a" it tries to parse a, as
it is possible to print expressions etc, and when is parsing in, in
exp_c.y, it throws errors if the variable is not in scope.

So currently my code sets a global variable check_scope, and in
exp_c.y, if check_scope is 1, then an internal variable is set if the
variable is not in scope, and an _error() is thrown, with an empty
string, and when running a script and an error is encountered, if the
error message is the empty string then the error is ignored.

Obviously this is not a particularly neat way of doing it, and I'm
sure there is a better way, so any help is appreciated.

Thanks
Rob

Re: Checking variable scope

[Jim Blandy]

"Rob Quill" <rob.quill@gmail.com> writes:
> The aim would be to add a command to gdb, in_scope (or similar) which
> can be used from a GDB script to check if a variable is in scope
> without throwing an error if is isn't. As currently if I do "print a"
> and a is not in scope then I get an error and the script stops, which
> is fine. But even better would be if I were able to check if a was in
> scope and only print (or do other things with it) if it was.

Why don't you add a new expression operator, $in_scope(a), which
evaluates to zero or one?  You could just imitate the other code in
the parser and expression evaluator, so this should be straightforward
to implement, and then you could use it in GDB script conditional
commands.

Checking variable scope

[Rob Quill]

On 28/11/06, Jim Blandy <jimb@codesourcery.com> wrote:
>
> "Rob Quill" <rob.quill@gmail.com> writes:
> > The aim would be to add a command to gdb, in_scope (or similar) which
> > can be used from a GDB script to check if a variable is in scope
> > without throwing an error if is isn't. As currently if I do "print a"
> > and a is not in scope then I get an error and the script stops, which
> > is fine. But even better would be if I were able to check if a was in
> > scope and only print (or do other things with it) if it was.
>
> Why don't you add a new expression operator, $in_scope(a), which
> evaluates to zero or one?  You could just imitate the other code in
> the parser and expression evaluator, so this should be straightforward
> to implement, and then you could use it in GDB script conditional
> commands.
>

Could you elaborate on this a bit please? I'm not really sure what you
mean and how to go about implementing it.

Thanks,
Rob

Re: Checking variable scope

[Jim Blandy]

"Rob Quill" <rob.quill@gmail.com> writes:
> On 28/11/06, Jim Blandy <jimb@codesourcery.com> wrote:
>>
>> "Rob Quill" <rob.quill@gmail.com> writes:
>> > The aim would be to add a command to gdb, in_scope (or similar) which
>> > can be used from a GDB script to check if a variable is in scope
>> > without throwing an error if is isn't. As currently if I do "print a"
>> > and a is not in scope then I get an error and the script stops, which
>> > is fine. But even better would be if I were able to check if a was in
>> > scope and only print (or do other things with it) if it was.
>>
>> Why don't you add a new expression operator, $in_scope(a), which
>> evaluates to zero or one?  You could just imitate the other code in
>> the parser and expression evaluator, so this should be straightforward
>> to implement, and then you could use it in GDB script conditional
>> commands.
>>
>
> Could you elaborate on this a bit please? I'm not really sure what you
> mean and how to go about implementing it.

Right now, c-exp.y parses a C expression into an expression,
represented using the structure in expression.h, and evaluated by the
code in eval.c.  But c-exp.y actually recognizes an extended grammar,
with special additions for debugging, like the @ operator and array
literals.

I'm suggesting that you extend the grammar even further, by allowing
one to say things like this (assume there is no binding for 'a' in
scope, but there is a binding for 'b' in scope):

(gdb) print $in_scope(a)
$1 = 0
(gdb) print $in_scope(b)
$1 = 1
(gdb)

Re: Checking variable scope

[Daniel Jacobowitz]

On Wed, Nov 29, 2006 at 10:44:33AM -0800, Jim Blandy wrote:
> I'm suggesting that you extend the grammar even further, by allowing
> one to say things like this (assume there is no binding for 'a' in
> scope, but there is a binding for 'b' in scope):
> 
> (gdb) print $in_scope(a)
> $1 = 0
> (gdb) print $in_scope(b)
> $1 = 1
> (gdb)

Is extending the grammar really necessary?  That will already parse;
it's the same as you'd use for a convenience variable named in_scope
holding a function pointer.

-- 
Daniel Jacobowitz
CodeSourcery

Re: Checking variable scope

[Jim Blandy]

Daniel Jacobowitz <drow@false.org> writes:
> On Wed, Nov 29, 2006 at 10:44:33AM -0800, Jim Blandy wrote:
>> I'm suggesting that you extend the grammar even further, by allowing
>> one to say things like this (assume there is no binding for 'a' in
>> scope, but there is a binding for 'b' in scope):
>> 
>> (gdb) print $in_scope(a)
>> $1 = 0
>> (gdb) print $in_scope(b)
>> $1 = 1
>> (gdb)
>
> Is extending the grammar really necessary?  That will already parse;
> it's the same as you'd use for a convenience variable named in_scope
> holding a function pointer.

I think so, because you need to prevent the reference to 'a' from
producing an error when you parse the expression.