bug#25561: 25.1; "Making isearch-search-fun-function local to *Minibuf-1* while let-bound!" warning with Evil

bug#25561: 25.1; "Making isearch-search-fun-function local to *Minibuf-1* while let-bound!" warning with Evil

[Vasilij Schneidermann]

Hello,

a user of the Evil package reported [1] that using a search command
results in a "Making isearch-search-fun-function local to *Minibuf-1*
while let-bound!" warning.  This appears to be an interaction between
isearch and Evil, more specifically, Evil let-binds
`isearch-search-fun-function' to something of its own, invokes isearch
and something in isearch ends up running
`minibuffer-history-isearch-setup' which `make-local-variable' to have
`isearch-search-fun-function' have a separate value, resulting in the
aforementioned warning.

What's the recommended way to avoid or silence it?  Stefan suggested
elsewhere [2] that this might indicate a bug, hence why I'm bringing
this up here.

[1]: https://github.com/emacs-evil/evil/issues/748
[2]: https://lists.gnu.org/archive/html/emacs-devel/2010-06/msg00903.html

bug#25561: 25.1; "Making isearch-search-fun-function local to *Minibuf-1* while let-bound!" warning with Evil

[Vasilij Schneidermann]

> Thanks for the report.  I'm not familiar with the Evil package, so
> please correct me if I make wrong guesses about it.  I suppose that
> the basic mode of operation of the Evil package is not buffer-local.

It's a bit trickier than that.  Evil defines a minor mode
(evil-local-mode), then a global minor mode (evil-mode) using it and a
few more variables for state tracking.  The supported mode of operation
is therefore global, while still allowing to deactivate it on a
per-buffer or even per-command basis (there's a command to deactivate
Evil's keymaps for the duration of the next command).

> OTOH, if evil-mode is the global mode then it makes sense to change
> the global value of isearch-search-fun-function (and perhaps keeping
> its old value).  If you need to keep standard search commands working,
> then you could create a new internal defvar e.g.
> evil-search-is-active, set it temporarily in evil-search commands, and
> check it in your new implementation of the search function in
> isearch-search-fun-function.  If evil-search-is-active is nil, then
> call the old default value isearch-search-fun-default.

Hm, this sounds hackier than the current approach, also, what if another
package let-binds this variable for custom search functionality?  The
reason I reported this warning is because I suspect something about
isearch and the minibuffer setup hook isn't done correctly, in hope of
getting rid of the warning for every user of that variable.

bug#25561: 25.1; "Making isearch-search-fun-function local to *Minibuf-1* while let-bound!" warning with Evil

[Vasilij Schneidermann]

Bumping this thread to CC Stefan Monnier as well.

On 01/29/17 at 02:39am, Juri Linkov wrote:
> > Hello,
> >
> > a user of the Evil package reported [1] that using a search command
> > results in a "Making isearch-search-fun-function local to *Minibuf-1*
> > while let-bound!" warning.  This appears to be an interaction between
> > isearch and Evil, more specifically, Evil let-binds
> > `isearch-search-fun-function' to something of its own, invokes isearch
> > and something in isearch ends up running
> > `minibuffer-history-isearch-setup' which `make-local-variable' to have
> > `isearch-search-fun-function' have a separate value, resulting in the
> > aforementioned warning.
> >
> > What's the recommended way to avoid or silence it?  Stefan suggested
> > elsewhere [2] that this might indicate a bug, hence why I'm bringing
> > this up here.
> >
> > [1]: https://github.com/emacs-evil/evil/issues/748
> > [2]: https://lists.gnu.org/archive/html/emacs-devel/2010-06/msg00903.html
> 
> Thanks for the report.  I'm not familiar with the Evil package,
> so please correct me if I make wrong guesses about it.  I suppose
> that the basic mode of operation of the Evil package is not buffer-local.
> In this case I believe (make-local-variable 'isearch-search-fun-function)
> is the wrong fix because it will make this variable have this value
> in every random buffer where the user happens to invoke the search command.
> 
> OTOH, if evil-mode is the global mode then it makes sense to change the
> global value of isearch-search-fun-function (and perhaps keeping its
> old value).  If you need to keep standard search commands working,
> then you could create a new internal defvar e.g. evil-search-is-active,
> set it temporarily in evil-search commands, and check it in your new
> implementation of the search function in isearch-search-fun-function.
> If evil-search-is-active is nil, then call the old default value
> isearch-search-fun-default.
> 
> I can't guarantee this solution to work, it's just an idea for you
> to consider.

bug#25561: 25.1; "Making isearch-search-fun-function local to *Minibuf-1* while let-bound!" warning with Evil

[Stefan Monnier]

>> > a user of the Evil package reported [1] that using a search command
>> > results in a "Making isearch-search-fun-function local to *Minibuf-1*
>> > while let-bound!" warning.  This appears to be an interaction between
>> > isearch and Evil, more specifically, Evil let-binds
>> > `isearch-search-fun-function' to something of its own, invokes isearch
>> > and something in isearch ends up running
>> > `minibuffer-history-isearch-setup' which `make-local-variable' to have
>> > `isearch-search-fun-function' have a separate value, resulting in the
>> > aforementioned warning.
>> >
>> > What's the recommended way to avoid or silence it?  Stefan suggested
>> > elsewhere [2] that this might indicate a bug, hence why I'm bringing
>> > this up here.

I plead guilty of adding this warning.  But FWIW, I think it was
misguided: the interaction between make-local-variable and let-binding
can be pretty tricky, which is why I added the warning (thinking that
such things should better be avoided), but in my subsequent experience,
I haven't found a single case where the problem is real.  Worse: it's
often difficult/ugly to change the code to avoid the "tricky"
interaction, because the code that does the let-binding doesn't know
about the other and vice-versa.

So, while the trickiness rmains, I think it's best to just hope for the
best, silently, rather than signal a warning that nobody (not even me)
knows how to address satisfactorily.


        Stefan

bug#25561: 25.1; "Making isearch-search-fun-function local to *Minibuf-1* while let-bound!" warning with Evil

[Eli Zaretskii]

> From: Stefan Monnier <monnier@IRO.UMontreal.CA>
> Date: Wed, 08 Feb 2017 17:21:11 -0500
> Cc: 25561@debbugs.gnu.org, Juri Linkov <juri@linkov.net>
> 
> >> > a user of the Evil package reported [1] that using a search command
> >> > results in a "Making isearch-search-fun-function local to *Minibuf-1*
> >> > while let-bound!" warning.  This appears to be an interaction between
> >> > isearch and Evil, more specifically, Evil let-binds
> >> > `isearch-search-fun-function' to something of its own, invokes isearch
> >> > and something in isearch ends up running
> >> > `minibuffer-history-isearch-setup' which `make-local-variable' to have
> >> > `isearch-search-fun-function' have a separate value, resulting in the
> >> > aforementioned warning.
> >> >
> >> > What's the recommended way to avoid or silence it?  Stefan suggested
> >> > elsewhere [2] that this might indicate a bug, hence why I'm bringing
> >> > this up here.
> 
> I plead guilty of adding this warning.  But FWIW, I think it was
> misguided: the interaction between make-local-variable and let-binding
> can be pretty tricky, which is why I added the warning (thinking that
> such things should better be avoided), but in my subsequent experience,
> I haven't found a single case where the problem is real.  Worse: it's
> often difficult/ugly to change the code to avoid the "tricky"
> interaction, because the code that does the let-binding doesn't know
> about the other and vice-versa.
> 
> So, while the trickiness rmains, I think it's best to just hope for the
> best, silently, rather than signal a warning that nobody (not even me)
> knows how to address satisfactorily.

So you are saying we should remove that warning entirely?

What about the other warning, "Making FOO buffer-local while let-bound!"?

bug#25561: 25.1; "Making isearch-search-fun-function local to *Minibuf-1* while let-bound!" warning with Evil

[Stefan Monnier]

> So you are saying we should remove that warning entirely?

Yes.

> What about the other warning, "Making FOO buffer-local while let-bound!"?

IIRC this one is less annoying, but it didn't prove useful either, so
I thing we should remove it as well.


        Stefan

bug#25561: 25.1; "Making isearch-search-fun-function local to *Minibuf-1* while let-bound!" warning with Evil

[Eli Zaretskii]

> From: Stefan Monnier <monnier@IRO.UMontreal.CA>
> Cc: v.schneidermann@gmail.com, 25561@debbugs.gnu.org, juri@linkov.net
> Date: Fri, 10 Feb 2017 07:45:53 -0500
> 
> > So you are saying we should remove that warning entirely?
> 
> Yes.
> 
> > What about the other warning, "Making FOO buffer-local while let-bound!"?
> 
> IIRC this one is less annoying, but it didn't prove useful either, so
> I thing we should remove it as well.

Done.