edebug-spec of setf: why is the getter evaluated on set?

edebug-spec of setf: why is the getter evaluated on set?

[Bruno BARBIER]

Hi,

I'm using the gv package (and setf). I really like it.

But, I don't understand why the GETTER is evaluated, when edebugging a
'setf' expression.

For example, let's take that expression:

    #+begin_src elisp
    (setf my:undefined 7)
    #+end_src

and let's assume the symbol "my:undefined" doesn't exist yet (that is,
"not interned").


Evaluating that expression with edebug raises the error:

    Symbol's value as variable is void: my:undefined


Evaluating that same expression outside of edebug, just sets the
variable, as expected.


In other words, the debug spec of 'setf' currently is:

    (&rest [gv-place form])

shouldn't it be something like:

    (&rest [sexp form])

or is it a bug in how 'gv-place' is handled ?

Thanks,

Bruno.


PS: If it matters, I'm using GNU Emacs 27.2

Re: edebug-spec of setf: why is the getter evaluated on set?

[Stefan Monnier]

> Evaluating that expression with edebug raises the error:
>
>     Symbol's value as variable is void: my:undefined
>
> Evaluating that same expression outside of edebug, just sets the
> variable, as expected.

Please report this as a bug.
>
> In other words, the debug spec of 'setf' currently is:
>
>     (&rest [gv-place form])
>
> shouldn't it be something like:
>
>     (&rest [sexp form])
>
> or is it a bug in how 'gv-place' is handled ?

We'll see,


        Stefan

Re: edebug-spec of setf: why is the getter evaluated on set?

[Bruno Barbier]

Hi Stefan,


Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> Evaluating that expression with edebug raises the error:
>>
>>     Symbol's value as variable is void: my:undefined
>>
>> Evaluating that same expression outside of edebug, just sets the
>> variable, as expected.
>
> Please report this as a bug.

You've already fixed it a while ago ... :-)

It seems it is bug 46573. And, it's not in emacs 27 (my mistake).

Sorry for the noise.


Bruno


>>
>> In other words, the debug spec of 'setf' currently is:
>>
>>     (&rest [gv-place form])
>>
>> shouldn't it be something like:
>>
>>     (&rest [sexp form])
>>
>> or is it a bug in how 'gv-place' is handled ?
>
> We'll see,
>
>
>         Stefan