Warning compiling obarray.el on master

Warning compiling obarray.el on master

[Eli Zaretskii]

    ELC      ../lisp/obarray.elc

  In toplevel form:
  obarray.el:40:1:Warning: Lexical argument shadows the dynamic variable obarray

Re: Warning compiling obarray.el on master

[Stefan Monnier]

>     ELC      ../lisp/obarray.elc
>   In toplevel form:
>   obarray.el:40:1:Warning: Lexical argument shadows the dynamic variable obarray

It's "harmless" and it's the result of having a global variable with
a short name like `obarray` which hence unavoidably collides with
local uses.

I see the rest of obarray.el uses the name `ob` for local variables
instead, so I guess using that in `obarray-copy` is the best fix in the
short term.
In the longer term, maybe we should rename `obarray` to something more
descriptive and less likely to collide with a local var.
Maybe `default-obarray` or `global-obarray` or
`global-identifier-obarray`?


        Stefan

Re: Warning compiling obarray.el on master

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Date: Sat, 18 Mar 2017 08:49:05 -0400
> 
> I see the rest of obarray.el uses the name `ob` for local variables
> instead, so I guess using that in `obarray-copy` is the best fix in the
> short term.

Please do.

> In the longer term, maybe we should rename `obarray` to something more
> descriptive and less likely to collide with a local var.
> Maybe `default-obarray` or `global-obarray` or
> `global-identifier-obarray`?

Perhaps, but the danger of clashing names doesn't go away with these
measures, it just gets lower.

Re: Warning compiling obarray.el on master

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

>> In the longer term, maybe we should rename `obarray` to something more
>> descriptive and less likely to collide with a local var.
>> Maybe `default-obarray` or `global-obarray` or
>> `global-identifier-obarray`?
>
> Perhaps, but the danger of clashing names doesn't go away with these
> measures, it just gets lower.

And not using the optional OBARRAY parameter in `intern' was marked as
obsolete a few years ago.  I think that was a mistake, but changing all
the code that says (intern "foo" obarray) now to
(intern "foo" global-identifier-obarray) wouldn't be, er, optimal, either.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Warning compiling obarray.el on master

[Noam Postavsky]

On Sat, Mar 18, 2017 at 9:47 AM, Lars Ingebrigtsen <larsi@gnus.org> wrote:
>
> And not using the optional OBARRAY parameter in `intern' was marked as
> obsolete a few years ago.  I think that was a mistake, but changing all
> the code that says (intern "foo" obarray) now to
> (intern "foo" global-identifier-obarray) wouldn't be, er, optimal, either.

I think you are mixing up intern (for which the optional OBARRAY is
not obsolete), with unintern which is much more rare. A quick grep
finds only 1 use of unintern with the global 'obarray':

./obsolete/old-whitespace.el:780:
  (if (unintern "whitespace-unload-hook" obarray)

Re: Warning compiling obarray.el on master

[Lars Ingebrigtsen]

Noam Postavsky <npostavs@users.sourceforge.net> writes:

> I think you are mixing up intern (for which the optional OBARRAY is
> not obsolete), with unintern which is much more rare. A quick grep
> finds only 1 use of unintern with the global 'obarray':

Oh, sorry.  I misremembered.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Warning compiling obarray.el on master

[Stefan Monnier]

> Perhaps, but the danger of clashing names doesn't go away with these
> measures, it just gets lower.

True, but that applies to most/all global variables we have.
And it's important to take into account that the reason why the danger
is reduced is not just because "longer names are more rare and contain
more entropy" or something like that, but because the longer name better
describes the variable, so it reduces the risk that someone else will
use that same identifier to refer to something different.


        Stefan