bug#3465: 23.0.94; feature request: be able to log minibuffer messages

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Drew Adams]

Feature request: Be able to log minibuffer messages, that is, messages
from `minibuffer-message'.
 
For example, option `message-log-max', which allows for disabling
logging, could have additional values:
 
* negative N means (abs N) max lines + log minibuf msgs too
* other non-nil value, besides t and N, means log minibuf msgs too
 
With minibuffer messages logged also, a user can consult *Messages*
(e.g. using `C-x C-b') if s?he misses an ephemeral message.

In GNU Emacs 23.0.94.1 (i386-mingw-nt5.1.2600)
 of 2009-05-24 on SOFT-MJASON
Windowing system distributor `Microsoft Corp.', version 5.1.2600
configured using `configure --with-gcc (3.4)'
 

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Lars Ingebrigtsen]

"Drew Adams" <drew.adams@oracle.com> writes:

> Feature request: Be able to log minibuffer messages, that is, messages
> from `minibuffer-message'.

(progn (minibuffer-message "foo") (message "zot"))

makes "foo" and "zot" end up in the *Message* buffer, so this seems like
it has been fixed already.

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

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Drew Adams]

> > Feature request: Be able to log minibuffer messages, that is, messages
> > from `minibuffer-message'.
> 
> (progn (minibuffer-message "foo") (message "zot"))
> 
> makes "foo" and "zot" end up in the *Message* buffer, so this seems like
> it has been fixed already.

OK, thanks.  But did you see the other part of the report, about
being able to use `message-log-max' to control minibuffer messages?

(Not critical.)

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Johan Bockgård]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> "Drew Adams" <drew.adams@oracle.com> writes:
>
>> Feature request: Be able to log minibuffer messages, that is, messages
>> from `minibuffer-message'.
>
> (progn (minibuffer-message "foo") (message "zot"))

>
> makes "foo" and "zot" end up in the *Message* buffer, so this seems like
> it has been fixed already.

You are supposed to test `minibuffer-message' when the current buffer is
a minibuffer. Then it doesn't log, so this is not fixed.

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Lars Ingebrigtsen]

Johan Bockgård <bojohan@gnu.org> writes:

> You are supposed to test `minibuffer-message' when the current buffer is
> a minibuffer. Then it doesn't log, so this is not fixed.

Ah, I see.

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

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Lars Ingebrigtsen]

A test case is

(with-current-buffer (get-buffer " *Minibuf-1*") (minibuffer-message "foo"))

and look in the *Messages* buffer.

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

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Lars Ingebrigtsen]

Lars Ingebrigtsen <larsi@gnus.org> writes:

> A test case is
>
> (with-current-buffer (get-buffer " *Minibuf-1*") (minibuffer-message "foo"))
>
> and look in the *Messages* buffer.

The problem here is that "foo" doesn't end up in *Messages*, and it
seems like it should.

I looked around for a utility function to add stuff to the *Messages*
buffer (it has to heed the message-log-max variable etc), but I couldn't
find one.

Would it make sense to add one?  And then make minibuffer-message (in
the use case above) use it?  Should be trivial to add, but I'm just
wondering whether there was a reason this hadn't been added already...

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

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Date: Thu, 20 Aug 2020 20:52:36 +0200
> Cc: 3465@debbugs.gnu.org
> 
> I looked around for a utility function to add stuff to the *Messages*
> buffer (it has to heed the message-log-max variable etc), but I couldn't
> find one.

??? What's wrong with 'message'?

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

>> I looked around for a utility function to add stuff to the *Messages*
>> buffer (it has to heed the message-log-max variable etc), but I couldn't
>> find one.
>
> ??? What's wrong with 'message'?

minibuffer-message can't use message here, apparently.

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

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: bojohan@gnu.org,  3465@debbugs.gnu.org
> Date: Thu, 20 Aug 2020 21:18:23 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> >> I looked around for a utility function to add stuff to the *Messages*
> >> buffer (it has to heed the message-log-max variable etc), but I couldn't
> >> find one.
> >
> > ??? What's wrong with 'message'?
> 
> minibuffer-message can't use message here, apparently.

I'm probably missing something.  This "works for me"™:

  M-: (with-current-buffer (get-buffer " *Minibuf-1*") (minibuffer-message "foo") (message "FOO")) RET

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> I'm probably missing something.  This "works for me"™:
>
>   M-: (with-current-buffer (get-buffer " *Minibuf-1*")
> (minibuffer-message "foo") (message "FOO")) RET

Does both foo and FOO end up in your *Messages* buffer?  I only get FOO
there, and the bug report is about the foo.

Foo fightin'.

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

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: bojohan@gnu.org,  3465@debbugs.gnu.org
> Date: Thu, 20 Aug 2020 21:45:24 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > I'm probably missing something.  This "works for me"™:
> >
> >   M-: (with-current-buffer (get-buffer " *Minibuf-1*")
> > (minibuffer-message "foo") (message "FOO")) RET
> 
> Does both foo and FOO end up in your *Messages* buffer?  I only get FOO
> there, and the bug report is about the foo.

I don't understand: I thought the above shows that calling 'message'
from the minibuffer does work, so programs which want to log
minibuffer messages can call 'message' with whatever they want to log,
and that stuff will end up in *Messages*.

What am I missing?

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

> I don't understand: I thought the above shows that calling 'message'
> from the minibuffer does work, so programs which want to log
> minibuffer messages can call 'message' with whatever they want to log,
> and that stuff will end up in *Messages*.
>
> What am I missing?

There's no problem with logging in general.  The request is to also have
the messages output by the minibuffer-message function also end up in
the *Messages* buffer.

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

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Juri Linkov]

> There's no problem with logging in general.  The request is to also have
> the messages output by the minibuffer-message function also end up in
> the *Messages* buffer.

This was already implemented in 8693611136 (see "Record message in the *Messages* buffer"
in 'minibuffer-message'), but later reverted in 7254b63462.

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Lars Ingebrigtsen]

Juri Linkov <juri@linkov.net> writes:

>> There's no problem with logging in general.  The request is to also have
>> the messages output by the minibuffer-message function also end up in
>> the *Messages* buffer.
>
> This was already implemented in 8693611136 (see "Record message in the
> *Messages* buffer"
> in 'minibuffer-message'), but later reverted in 7254b63462.

The reversion refers to

    Revert aa89c84e00d8dc85100e6fedab7631c415e6364d (bug#38457)

which is a very long thread.  What was the reason for removing the
*Messages* logging?  The patch had other things in addition to the
logging...

But I liked the way to log to the buffer: 

+    ;; Record message in the *Messages* buffer
+    (let ((inhibit-message t))
+      (if args
+          (apply #'message message args)
+        (message "%s" message)))


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

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Eli Zaretskii]

> From: Lars Ingebrigtsen <larsi@gnus.org>
> Cc: bojohan@gnu.org,  3465@debbugs.gnu.org
> Date: Thu, 20 Aug 2020 22:03:28 +0200
> 
> There's no problem with logging in general.  The request is to also have
> the messages output by the minibuffer-message function also end up in
> the *Messages* buffer.

By default?  I'd object to that.  As an option, maybe; but then I
don't understand why this is requested as a core feature, since any
Lisp program that wants this can simply call 'message' with the same
string, and be done.

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Lars Ingebrigtsen]

Eli Zaretskii <eliz@gnu.org> writes:

>> There's no problem with logging in general.  The request is to also have
>> the messages output by the minibuffer-message function also end up in
>> the *Messages* buffer.
>
> By default?  I'd object to that.  As an option, maybe; but then I
> don't understand why this is requested as a core feature, since any
> Lisp program that wants this can simply call 'message' with the same
> string, and be done.

I thought it would make sense to have that function behave as regularly
as possible...

However, looking at the use cases for minibuffer-message, all of the
in-tree calls are trivial: "Confirm".  "Incomplete".  "Hit space to
flush".  All of those things are UI details that would be surprising to
have land in the *Messages* buffer.

So I agree with Eli, and I don't think this is something that
minibuffer-message should do at all.  If something wants to do a
minibuffer-message and log it, then that thing should just do both.

So I'm closing this bug report.

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

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Drew Adams]

> >> There's no problem with logging in general.  The request is to also have
> >> the messages output by the minibuffer-message function also end up in
> >> the *Messages* buffer.

No.  The request is to _be able_ to also log
`minibuffer-message' feedback.

> > By default?  I'd object to that.  As an option, maybe; 

The bug requested to _be able to_ do it.
Not to do it by default.

> > but then I
> > don't understand why this is requested as a core feature, since any
> > Lisp program that wants this can simply call 'message' with the same
> > string, and be done.

No.  `minibuffer-message' has an entirely different
UI effect.  You are coupling the UI effect (messaging)
with the logging effect.  It is you, even in your "but
then" (i.e., NOT by default), who are saying that
every time logging is wanted the UI effect of `message'
is also wanted.

That's exactly the point of this bug:

1. `message' and `minibuffer-message' have different UI
   effects.  They are used for different things.

2. `message' can be logged. `minibuffer-message' cannot.

The point is to be able to log `minibuffer-message',
WITHOUT getting the different UI effect of `message'.

Please reread the bug report.  I don't think it's hard
to understand the request: Be able to also log
`minibuffer-message' output.

And obviously WITHOUT calling `message', which is not
logging `minibuffer-message' but logging `message',
and which adds the UI effect of `message'.

> I thought it would make sense to have that function behave as regularly
> as possible...
> 
> However, looking at the use cases for minibuffer-message, all of the
> in-tree calls are trivial: "Confirm".  "Incomplete".  "Hit space to
> flush".  All of those things are UI details that would be surprising to
> have land in the *Messages* buffer.

This is irrelevant.  User code - 3rd-party libraries
can, and do, use `minibuffer-message' in other ways.

Don't judge what something is used for generally by
such trivial uses of it.  Uses of the minibuffer by
vanilla Emacs are mostly trivial,  There's not much
user interaction, not many keys/actions possible in
the minibuffer, etc.

> So I agree with Eli, and I don't think this is something that
> minibuffer-message should do at all.  If something wants to do a
> minibuffer-message and log it, then that thing should just do both.
> 
> So I'm closing this bug report.

Eli seems to have clearly misunderstood the request.

And you seem to have based your conclusion of how
`minibuffer-message' is and can be used only on its
few uses in the vanilla code, which are trivial uses.

Please reopen this enhancement request.  It's simply
a request to _be able to_ log `minibuffer-message'
output.  Not a big deal.

`message' does NOT replace `minibuffer-message', and
vice versa.  Neither should be hard-code coupled with
logging.  And each should be able to log, as well as
have its minibuffer UI effect.

The effects of the two functions should be totally
_independent_.

(And yes, Juri's recent changes that act against this
independence are a real step _backward_.  I argued
this in vain in that context.  Too bad.)

bug#4477: bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Eli Zaretskii]

> Date: Sat, 22 Aug 2020 09:54:14 -0700 (PDT)
> From: Drew Adams <drew.adams@oracle.com>
> Cc: 3465@debbugs.gnu.org, bojohan@gnu.org, 4477@debbugs.gnu.org
> 
> > >> There's no problem with logging in general.  The request is to also have
> > >> the messages output by the minibuffer-message function also end up in
> > >> the *Messages* buffer.
> 
> No.  The request is to _be able_ to also log
> `minibuffer-message' feedback.
> 
> > > By default?  I'd object to that.  As an option, maybe; 
> 
> The bug requested to _be able to_ do it.
> Not to do it by default.

I was responding to Lars, not to your original request.

> > > but then I
> > > don't understand why this is requested as a core feature, since any
> > > Lisp program that wants this can simply call 'message' with the same
> > > string, and be done.
> 
> No.  `minibuffer-message' has an entirely different
> UI effect.  You are coupling the UI effect (messaging)
> with the logging effect.  It is you, even in your "but
> then" (i.e., NOT by default), who are saying that
> every time logging is wanted the UI effect of `message'
> is also wanted.

Yes, yes, everyone knows that.  No need to reiterate the trivia.
Please focus on the real issues at hand.

> That's exactly the point of this bug:
> 
> 1. `message' and `minibuffer-message' have different UI
>    effects.  They are used for different things.
> 
> 2. `message' can be logged. `minibuffer-message' cannot.
> 
> The point is to be able to log `minibuffer-message',
> WITHOUT getting the different UI effect of `message'.

Yes, we understood that.  No need to reiterate.

> Please reread the bug report.  I don't think it's hard
> to understand the request: Be able to also log
> `minibuffer-message' output.

Please re-read what I responded.  Which part of it contradicts what
you requested?

> And obviously WITHOUT calling `message', which is not
> logging `minibuffer-message' but logging `message',
> and which adds the UI effect of `message'.

The UI effect of 'message' can be easily disabled by the Lisp program
which calls 'message', which in effect converts 'message' into a
logging function.  Include that in the code that calls
'minibuffer-message', and you have what you wanted, no?

> Eli seems to have clearly misunderstood the request.

Please don't second-guess my misunderstandings.  It is rude, to say
the least.

> Please reopen this enhancement request.  It's simply
> a request to _be able to_ log `minibuffer-message'
> output.  Not a big deal.

You are already able to do that, AFAICT.

bug#4477: bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Drew Adams]

> > Please reread the bug report.  I don't think it's hard
> > to understand the request: Be able to also log
> > `minibuffer-message' output.
> 
> Please re-read what I responded.  Which part of it
> contradicts what you requested?
> 
> > And obviously WITHOUT calling `message', which is not
> > logging `minibuffer-message' but logging `message',
> > and which adds the UI effect of `message'.
> 
> The UI effect of 'message' can be easily disabled by the Lisp program
> which calls 'message', which in effect converts 'message' into a
> logging function.  Include that in the code that calls
> 'minibuffer-message', and you have what you wanted, no?

Please read what you just responded to again.

The request is to be able to have `minibuffer-message'
output logged, WITHOUT calling `message'.

You seem to keep repeating that we can have the effect
of logging the text (and have the UI effect of only
`minibuffer-message') by ALSO calling `message'.

That's not the request.  It's not about whether one
can get the effect of logging `minibuffer-message'
output without also logging a `message' output for
the same text.

It's about simply controlling logging for
`minibuffer-message', totally, completely, independent
of any use of `message'.

IOW, please forget about `message'.  The request is
for a simple way to (optionally) log output of 
`minibuffer-message', just as we do, for example for
`message' output.  And without recourse to any call
to `message' - no workaround, just a simple way to
log `minibuffer-message'.  We have such a way for
`message' output.  The request is for such a way for
`minibuffer-message' output - totally independent
from `message'.

> > Eli seems to have clearly misunderstood the request.
> 
> Please don't second-guess my misunderstandings.  It is rude, to say
> the least.

Fair enough to say that I do think you've misundertood?

Perhaps you prefer RMS's favorite way to put it: "We
seem to be miscommunicating."

You reply that what I say is obvious to all and doesn't
need to be said.  I reply that I think you're not
getting what I'm saying.  You reply that there's a
workaround (which I could say is obvious to all and
doesn't need to be reiterated).

I reply that I'm aware of that workaround, and I'm
asking for a _simple_ way to do what's requested,
along the lines of what's available for `message',
for instance.

> > Please reopen this enhancement request.  It's simply
> > a request to _be able to_ log `minibuffer-message'
> > output.  Not a big deal.
> 
> You are already able to do that, AFAICT.

Not simply.  Not in a way similar or analogous to
what we can do for `message'.   Not without jumping
through a hoop - also calling `message'.

Please keep `message' out of it.  This request is
about `minibuffer-message'.

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Lars Ingebrigtsen]

Drew Adams <drew.adams@oracle.com> writes:

> Please read what you just responded to again.
>
> The request is to be able to have `minibuffer-message'
> output logged, WITHOUT calling `message'.

We understand the request, and we have said "no, we're not going to add
that to Emacs".

If you want to do a minibuffer message and log it, you can write your
own function:

(defun minibuffer-message-and-log (&rest args)
  (let ((inhibit-message t))
    (apply #'message args))
  (apply #'minibuffer-message args))

Or whatever.

This seems to be a common thread when you request extending a function
to do something special in one way or another, when that can be done
trivially by just calling two functions instead.

And the rearguard-fighting action when these requests are rejected are
tiresome, to say the least.  Please stop doing it.

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

bug#4477: bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Eli Zaretskii]

> Date: Sat, 22 Aug 2020 11:11:23 -0700 (PDT)
> From: Drew Adams <drew.adams@oracle.com>
> Cc: larsi@gnus.org, 3465@debbugs.gnu.org, bojohan@gnu.org,
>         4477@debbugs.gnu.org
> 
> The request is to be able to have `minibuffer-message'
> output logged, WITHOUT calling `message'.

Why "without"?

> It's about simply controlling logging for
> `minibuffer-message', totally, completely, independent
> of any use of `message'.

You want every single call to minibuffer-message to be logged in
*Messages*?  I'm guessing not, because this would make no sense.

If you want to log only _some_ calls to minibuffer-message, then you
will have to have special code to request that in those places where
you want the minibuffer messages to be logged.  I'm saying that you
can easily call 'message' in those places, and be done with that.

> IOW, please forget about `message'.  The request is
> for a simple way to (optionally) log output of 
> `minibuffer-message', just as we do, for example for
> `message' output.  And without recourse to any call
> to `message' - no workaround, just a simple way to
> log `minibuffer-message'.  We have such a way for
> `message' output.  The request is for such a way for
> `minibuffer-message' output - totally independent
> from `message'.

You cannot request a feature and then also dictate how it is
implemented.  The solution I proposed should do what you want without
requiring any changes to the core.

> > You are already able to do that, AFAICT.
> 
> Not simply.

It's simple enough, from where I stand.  Certainly so for a minor
feature such as this one.

bug#3465: 23.0.94; feature request: be able to log minibuffer messages

[Drew Adams]

> > The request is to be able to have `minibuffer-message'
> > output logged, WITHOUT calling `message'.
> 
> Why "without"?

Treat `minibuffer-message' as first-class, the same
way `message' is treated.  `message' shouldn't affect
`minibuffer-message', and vice versa.

`message' has an easy way to do what's requested;
`minibuffer-message' does not have one.

> > It's about simply controlling logging for
> > `minibuffer-message', totally, completely, independent
> > of any use of `message'.
> 
> You want every single call to minibuffer-message to be logged in
> *Messages*?  I'm guessing not, because this would make no sense.

Not at all.  What makes you think that?

I think I made it clear that the request is to be able
to have, for `minibuffer-message', control over logging
similar to what we have for `message'.

They are separate.  They should remain separate.
All that's being requested is logging control
for `minibuffer-message' that's on a par with the
logging control for `message'.

> If you want to log only _some_ calls to minibuffer-message, then you
> will have to have special code to request that in those places where
> you want the minibuffer messages to be logged.  I'm saying that you
> can easily call 'message' in those places, and be done with that.

What I want is to be able to do with and for
`minibuffer-message' what I can do with and for
`message'.  Without any use of `message' - no
connection between the two, no dependence of
either on the other.

> > IOW, please forget about `message'.  The request is
> > for a simple way to (optionally) log output of
> > `minibuffer-message', just as we do, for example for
> > `message' output.  And without recourse to any call
> > to `message' - no workaround, just a simple way to
> > log `minibuffer-message'.  We have such a way for
> > `message' output.  The request is for such a way for
> > `minibuffer-message' output - totally independent
> > from `message'.
> 
> You cannot request a feature and then also dictate
> how it is implemented.

I in no way dictated anything about implementation.
I have said absolutely nothing about implementation.

The entire request is expressed in terms of behavior:
user-visible, user-controllable behavior.  Nothing
about implementation.  It's about what users can do.

> The solution I proposed should do what you want without
> requiring any changes to the core.

See above.  It doesn't do what was requested.

> > > You are already able to do that, AFAICT.
> >
> > Not simply.
> 
> It's simple enough, from where I stand.  Certainly
> so for a minor feature such as this one.