bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Gemini Lasswell]

The symbol &rest in an Edebug spec has a different meaning than it has
in an argument list. In particular, "&rest body" in an Edebug spec is
erroneous usage because it is asking Edebug to match something that
can match nothing (body) as many times as it can, which makes an
infinite loop.

To repro, navigate to the definition of crm-completion-help in
lisp/emacs-lisp/crm.el and:

M-x eval-buffer RET
C-u C-M-x

Result: Too deep - perhaps infinite loop in spec?

Here are the macros I've found that have this incorrect usage of &rest
in their Edebug specs:

crm--completion-command
easy-mmode-define-navigation
inline--letlisteval
inline--leteval
ses--letref

Edebug specs have been using &rest for a really long time and it
mostly gets used correctly. But &repeat would be better, since it
actually describes the functionality and doesn't have a completely
different and more common usage in argument lists.

Here's a little example of what an Edebug spec would look like with
&repeat in place of &rest:

(defmacro msgmac (&rest args)
  (declare (debug ("start" [&repeat "middle"] "end")))
  `(message "%s" ',args))

(defun use-msgmac ()
  (msgmac start end)
  (msgmac start middle end)
  (msgmac start middle middle end))

Does anyone have thoughts on the idea of adding &repeat as a synonym
for &rest in Edebug specs, with the goal of moving &rest towards
obsolescence eventually?

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Noam Postavsky]

Gemini Lasswell <gazally@runbox.com> writes:

> The symbol &rest in an Edebug spec has a different meaning than it has
> in an argument list. In particular, "&rest body" in an Edebug spec is
> erroneous usage because it is asking Edebug to match something that
> can match nothing (body) as many times as it can, which makes an
> infinite loop.

> Here are the macros I've found that have this incorrect usage of &rest
> in their Edebug specs:
>
> crm--completion-command
> easy-mmode-define-navigation
> inline--letlisteval
> inline--leteval
> ses--letref

> Does anyone have thoughts on the idea of adding &repeat as a synonym
> for &rest in Edebug specs, with the goal of moving &rest towards
> obsolescence eventually?

Makes sense to me, edebug specs can get pretty complicated.  Anything
which makes silly errors less likely is a win.

I wonder if would be worth trying to catch '&rest body' specs earlier
(ideally at compile time)?

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Lars Ingebrigtsen]

Gemini Lasswell <gazally@runbox.com> writes:

> The symbol &rest in an Edebug spec has a different meaning than it has
> in an argument list. In particular, "&rest body" in an Edebug spec is
> erroneous usage because it is asking Edebug to match something that
> can match nothing (body) as many times as it can, which makes an
> infinite loop.

[...]

> Here are the macros I've found that have this incorrect usage of &rest
> in their Edebug specs:
>
> crm--completion-command
> easy-mmode-define-navigation
> inline--letlisteval
> inline--leteval
> ses--letref

I've now fixed these.

> Edebug specs have been using &rest for a really long time and it
> mostly gets used correctly. But &repeat would be better, since it
> actually describes the functionality and doesn't have a completely
> different and more common usage in argument lists.
>
> Here's a little example of what an Edebug spec would look like with
> &repeat in place of &rest:
>
> (defmacro msgmac (&rest args)
>   (declare (debug ("start" [&repeat "middle"] "end")))
>   `(message "%s" ',args))
>
> (defun use-msgmac ()
>   (msgmac start end)
>   (msgmac start middle end)
>   (msgmac start middle middle end))
>
> Does anyone have thoughts on the idea of adding &repeat as a synonym
> for &rest in Edebug specs, with the goal of moving &rest towards
> obsolescence eventually?

I agree that this would be a lot clearer and probably less error-prone
than &rest.

Does anybody have an opinion here?

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

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > > The symbol &rest in an Edebug spec has a different meaning than it has
  > > in an argument list.

That seems like a confusing wrinkle.  Could we possibly change Emacs
to smooth it out?

-- 
Dr Richard Stallman
Founder, Free Software Foundation (https://gnu.org, https://fsf.org)
Internet Hall-of-Famer (https://internethalloffame.org)

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Lars Ingebrigtsen]

Richard Stallman <rms@gnu.org> writes:

>   > > The symbol &rest in an Edebug spec has a different meaning than it has
>   > > in an argument list.
>
> That seems like a confusing wrinkle.  Could we possibly change Emacs
> to smooth it out?

Yes, the suggestion was to make &repeat a synonym for &rest and then
replace all the usages.

I'm not sure this is worth the code churn, though -- as was pointed out,
virtually all the debug specs gets this right anyway.  Does anybody have
an opinion?

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

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > I'm not sure this is worth the code churn, though -- as was pointed out,
  > virtually all the debug specs gets this right anyway.  Does anybody have
  > an opinion?

The benefit would be to simplify the specs for Edebug specs.

The cost would be to fix the existing ones -- if they would need changing.
Would any existing correct ones need changing?

-- 
Dr Richard Stallman
Founder, Free Software Foundation (https://gnu.org, https://fsf.org)
Internet Hall-of-Famer (https://internethalloffame.org)

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Lars Ingebrigtsen]

Richard Stallman <rms@gnu.org> writes:

>   > I'm not sure this is worth the code churn, though -- as was pointed out,
>   > virtually all the debug specs gets this right anyway.  Does anybody have
>   > an opinion?
>
> The benefit would be to simplify the specs for Edebug specs.
>
> The cost would be to fix the existing ones -- if they would need changing.
> Would any existing correct ones need changing?

They wouldn't need changing, but it'd be confusing (for people reading
the code) to have some specs with &repeat and some with &rest, so I
think we should change them all if we decide to introduce the &repeat
alias.

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

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Michael Heerdegen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> I agree that this would be a lot clearer and probably less error-prone
> than &rest.
>
> Does anybody have an opinion here?

It could be worth changing it.  There are a lot of things where
understanding Edebug specs gets complicated.  Anything to make things
clearer is good.

Michael.

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Lars Ingebrigtsen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

>> Here's a little example of what an Edebug spec would look like with
>> &repeat in place of &rest:
>>
>> (defmacro msgmac (&rest args)
>>   (declare (debug ("start" [&repeat "middle"] "end")))
>>   `(message "%s" ',args))
>>
>> (defun use-msgmac ()
>>   (msgmac start end)
>>   (msgmac start middle end)
>>   (msgmac start middle middle end))
>>
>> Does anyone have thoughts on the idea of adding &repeat as a synonym
>> for &rest in Edebug specs, with the goal of moving &rest towards
>> obsolescence eventually?
>
> I agree that this would be a lot clearer and probably less error-prone
> than &rest.
>
> Does anybody have an opinion here?

Stefan added to the CCs; perhaps he has an opinion.

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

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Richard Stallman]

[[[ To any NSA and FBI agents reading my email: please consider    ]]]
[[[ whether defending the US Constitution against all enemies,     ]]]
[[[ foreign or domestic, requires you to follow Snowden's example. ]]]

  > Resent-From: Lars Ingebrigtsen <larsi@gnus.org>
  > Resent-CC: bug-gnu-emacs@gnu.org
  > Resent-Sender: help-debbugs@gnu.org
  > From: Lars Ingebrigtsen <larsi@gnus.org>
  > Date: Sat, 29 Jan 2022 16:56:03 +0100
  > Message-ID: <87o83u8rjg.fsf@gnus.org>

  > >> Here's a little example of what an Edebug spec would look like with
  > >> &repeat in place of &rest:
  > >>
  > >> (defmacro msgmac (&rest args)
  > >>   (declare (debug ("start" [&repeat "middle"] "end")))
  > >>   `(message "%s" ',args))
  > >>
  > >> (defun use-msgmac ()
  > >>   (msgmac start end)
  > >>   (msgmac start middle end)
  > >>   (msgmac start middle middle end))
  > >>
  > >> Does anyone have thoughts on the idea of adding &repeat as a synonym
  > >> for &rest in Edebug specs, with the goal of moving &rest towards
  > >> obsolescence eventually?

  ...

  > Stefan added to the CCs; perhaps he has an opinion.

It looks like this makes usage more complex, and the change would be
incompatible.  That is two reasons not to make the change.  Does it
offer any benefit?

The message I am replying to is clearly not the start of the thread,
but I can't find the earlier ones.  I searched my inboxes back to Oct 11.
It looks like somehow I did not receive them.

-- 
Dr Richard Stallman (https://stallman.org)
Chief GNUisance of the GNU Project (https://gnu.org)
Founder, Free Software Foundation (https://fsf.org)
Internet Hall-of-Famer (https://internethalloffame.org)

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors]

> The symbol &rest in an Edebug spec has a different meaning than it has
> in an argument list.

Definitely, yes.
Note that `&optional` is different "in the same way".

> In particular, "&rest body" in an Edebug spec is
> erroneous usage because it is asking Edebug to match something that
> can match nothing (body) as many times as it can, which makes an
> infinite loop.

Only idiots would get caught making such rookie mistakes.

> Here are the macros I've found that have this incorrect usage of &rest
> in their Edebug specs:
>
> crm--completion-command
> easy-mmode-define-navigation
> inline--letlisteval
> inline--leteval
> ses--letref

See: only idiots.

> Does anyone have thoughts on the idea of adding &repeat as a synonym
> for &rest in Edebug specs, with the goal of moving &rest towards
> obsolescence eventually?

I like the idea.  Only problem I can see is that it doesn't
solve the problem for `&optional`.


        Stefan

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Lars Ingebrigtsen]

Stefan Monnier via "Bug reports for GNU Emacs, the Swiss army knife of
text editors" <bug-gnu-emacs@gnu.org> writes:

>> Does anyone have thoughts on the idea of adding &repeat as a synonym
>> for &rest in Edebug specs, with the goal of moving &rest towards
>> obsolescence eventually?
>
> I like the idea.  Only problem I can see is that it doesn't
> solve the problem for `&optional`.

I guess it could also lead to problems in third-party packages that use
the edebug specs themselves -- they expect symbols tagged up with &rest
and get &repeat instead, breaking their code.

I'm not sure how serious a problem that is, though.

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

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors]

> I guess it could also lead to problems in third-party packages that use
> the edebug specs themselves -- they expect symbols tagged up with &rest
> and get &repeat instead, breaking their code.
> I'm not sure how serious a problem that is, though.

I'm not too worried about that.
[ And if we're wrong, it'd let us learn about uses of edebug specs of
  which we're not aware.  ]


        Stefan

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Lars Ingebrigtsen]

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

> I'm not too worried about that.
> [ And if we're wrong, it'd let us learn about uses of edebug specs of
>   which we're not aware.  ]

It's hard to grep for, but there's a usage in pp.el, at least, so it's
used outside of the edebug machinery.  

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

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Stefan Monnier via Bug reports for GNU Emacs, the Swiss army knife of text editors]

Lars Ingebrigtsen [2022-01-31 03:35:02] wrote:
> Stefan Monnier <monnier@iro.umontreal.ca> writes:
>> I'm not too worried about that.
>> [ And if we're wrong, it'd let us learn about uses of edebug specs of
>>   which we're not aware.  ]
> It's hard to grep for, but there's a usage in pp.el, at least, so it's
> used outside of the edebug machinery.

Yes, but it hardly qualifies for "of which we're not aware" ;-)


        Stefan

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Lars Ingebrigtsen]

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

>> It's hard to grep for, but there's a usage in pp.el, at least, so it's
>> used outside of the edebug machinery.
>
> Yes, but it hardly qualifies for "of which we're not aware" ;-)

No, but if one weirdo's made use of the specs, then there's probably
more out there.

Your point about &optional also makes me lean towards not doing anything
here -- I mean, "&optional body" and "&rest body" are both likely in
edebug specs, and the confusion here more stems from `body' in itself
being nullable, I think?

So I dunno whether renaming &rest to &repeat is going to help people
that write debug specs in practice much.

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

bug#28747: 26.0.60; Usage of "&rest body" in Edebug specs

[Lars Ingebrigtsen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> Your point about &optional also makes me lean towards not doing anything
> here -- I mean, "&optional body" and "&rest body" are both likely in
> edebug specs, and the confusion here more stems from `body' in itself
> being nullable, I think?
>
> So I dunno whether renaming &rest to &repeat is going to help people
> that write debug specs in practice much.

I think the conclusion here is that we should probably let sleeping dogs
doze, so I'm closing this bug report.

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