bug#6740: Spurious byte compiler warnings

bug#6740: Spurious byte compiler warnings

[Alan Mackenzie]

Hi, Emacs!

(i) Start Emacs, even a most recent bzr version, with -Q.
(ii) Put the following into the *scratch* buffer:

    (eval-when-compile
      (if (and (not (featurep 'cc-fix))
               (featurep 'xemacs)
               (progn
                 (require 'font-lock)
                 (let (font-lock-keywords)
                   (font-lock-compile-keywords '("\\<\\>"))
                   font-lock-keywords)))
          (cc-load "cc-fix")))

  (This fragment is at the top level, and taken from cc-defs.el.).
(iii) do M-x compile-defun on this form.

The byte compiler then issues the following two identical error
messages:

    Warning: value returned from (featurep (quote cc-fix)) is unused
    Warning: value returned from (featurep (quote cc-fix)) is unused

It is obvious that that value is indeed used.  This is a bug.

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#6740: Spurious byte compiler warnings

[Dan Nicolaescu]

Alan Mackenzie <acm@muc.de> writes:

> Hi, Emacs!
>
> (i) Start Emacs, even a most recent bzr version, with -Q.
> (ii) Put the following into the *scratch* buffer:
>
>     (eval-when-compile
>       (if (and (not (featurep 'cc-fix))
>                (featurep 'xemacs)
>                (progn
>                  (require 'font-lock)
>                  (let (font-lock-keywords)
>                    (font-lock-compile-keywords '("\\<\\>"))
>                    font-lock-keywords)))
>           (cc-load "cc-fix")))
>
>   (This fragment is at the top level, and taken from cc-defs.el.).
> (iii) do M-x compile-defun on this form.
>
> The byte compiler then issues the following two identical error
> messages:
>
>     Warning: value returned from (featurep (quote cc-fix)) is unused
>     Warning: value returned from (featurep (quote cc-fix)) is unused
>
> It is obvious that that value is indeed used.  This is a bug.

The byte compiler knows that (featurep 'xemacs) is false, so 
 (and (not (featurep 'cc-fix)) ... )
will be false, so the featurep result is indeed unused.

If you use (and (featurep 'xemacs) (not (featurep 'cc-fix) ...
the warning will go away.

bug#6740: Spurious byte compiler warnings

[Alan Mackenzie]

Hi, Dan,

On Tue, Jul 27, 2010 at 04:26:03PM -0400, Dan Nicolaescu wrote:
> Alan Mackenzie <acm@muc.de> writes:

> > Hi, Emacs!

> > (i) Start Emacs, even a most recent bzr version, with -Q.
> > (ii) Put the following into the *scratch* buffer:

> >     (eval-when-compile
> >       (if (and (not (featurep 'cc-fix))
> >                (featurep 'xemacs)
> >                (progn
> >                  (require 'font-lock)
> >                  (let (font-lock-keywords)
> >                    (font-lock-compile-keywords '("\\<\\>"))
> >                    font-lock-keywords)))
> >           (cc-load "cc-fix")))

> >   (This fragment is at the top level, and taken from cc-defs.el.).
> > (iii) do M-x compile-defun on this form.

> > The byte compiler then issues the following two identical error
> > messages:

> >     Warning: value returned from (featurep (quote cc-fix)) is unused
> >     Warning: value returned from (featurep (quote cc-fix)) is unused

> > It is obvious that that value is indeed used.  This is a bug.

> The byte compiler knows that (featurep 'xemacs) is false, so 
>  (and (not (featurep 'cc-fix)) ... )
> will be false, so the featurep result is indeed unused.

Ah, thanks for the explanation!  But .....

It's a bug that the error message is repeated, at the very least.  And
the message is most assuredly false because (featurep 'cc-fix) isn't
always false - for example when it's run under XEmacs.

> If you use (and (featurep 'xemacs) (not (featurep 'cc-fix) ...
> the warning will go away.

At the very least, emitting such arcane warnings is unhelpful.  Are there
any circumstances in which such a warning might help a hacker improve his
code?  Surely we aren't in the business of making it difficult to adapt
code for XEmacs?

I think it's clear, the only code containing (featurep 'xemacs) is
portable code.  Can we please remove this unhelpful warning?

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#6740: Spurious byte compiler warnings

[Juanma Barranquero]

On Tue, Jul 27, 2010 at 23:23, Alan Mackenzie <acm@muc.de> wrote:

> It's a bug that the error message is repeated, at the very least.

Yes, likely.

> And
> the message is most assuredly false because (featurep 'cc-fix) isn't
> always false - for example when it's run under XEmacs.

Not, in code byte-compiled for Emacs it is always false because
(featurep 'xemacs) is false. That means that optimized bytecode is not
portable to XEmacs. But I don't think it was before (for a long time).

> At the very least, emitting such arcane warnings is unhelpful.  Are there
> any circumstances in which such a warning might help a hacker improve his
> code?  Surely we aren't in the business of making it difficult to adapt
> code for XEmacs?

You make it appear as it if were an attempt to warn about using
XEmacs-specific code, but it is not, as Dan has pointed out. The
warning is generic, the result of

  (and X (featurep 'xemacs) Y Z...)  => (prog (and X) nil) => (prog X
nil)   ; IIUC the comments in byte-opt.el...

from whence, "value returned from X is unused".

> I think it's clear, the only code containing (featurep 'xemacs) is
> portable code.

The code is portable. The .elc is not.

    Juanma

bug#6740: Spurious byte compiler warnings

[Alan Mackenzie]

Hi, Juanma,

On Wed, Jul 28, 2010 at 12:57:04AM +0200, Juanma Barranquero wrote:
> On Tue, Jul 27, 2010 at 23:23, Alan Mackenzie <acm@muc.de> wrote:

> > It's a bug that the error message is repeated, at the very least.

> Yes, likely.

> > And the message is most assuredly false because (featurep 'cc-fix)
> > isn't always false - for example when it's run under XEmacs.

> Not, in code byte-compiled for Emacs it is always false because
> (featurep 'xemacs) is false. That means that optimized bytecode is not
> portable to XEmacs. But I don't think it was before (for a long time).

Optimised .elc isn't even compatible with earlier versions of Emacs.  But
this discussion is most emphatically NOT about compiled code.  It's about
the process of compiling sources.

> > At the very least, emitting such arcane warnings is unhelpful.  Are
> > there any circumstances in which such a warning might help a hacker
> > improve his code?  Surely we aren't in the business of making it
> > difficult to adapt code for XEmacs?

> You make it appear as it if were an attempt to warn about using
> XEmacs-specific code, but it is not, as Dan has pointed out.

What use is this warning message?  How could it prompt a hacker to
improve his code?  (That's a genuine question, not a rhetorical one.)

> The warning is generic, the result of

>   (and X (featurep 'xemacs) Y Z...)  => (prog (and X) nil) => (prog X
> nil)   ; IIUC the comments in byte-opt.el...

Thanks for this explanation.

What do you mean by "generic" here?  Is the same trick performed on
symbols other than 'xemacs?

Surely it is a bug that "(featurep 'cc-fix)" appears in the message
rather than "(featurep 'xemacs)".  This situation is not about 'cc-fix,
it's about 'xemacs.

I would suggest that if I, a highly experienced Emacs hacker, remain
baffled by this message over many months, so will lots of others.  Does
anybody actually care about "(featurep 'xemacs)" being optimised away?
Again, who does this warning message help (other than the hackers who
wrote the optimisation code in the first place, of course)?


> from whence, "value returned from X is unused".

> > I think it's clear, the only code containing (featurep 'xemacs) is
> > portable code.

> The code is portable. The .elc is not.

Er, we're in strong agreement with this.

>     Juanma

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#6740: Spurious byte compiler warnings

[Juanma Barranquero]

> But
> this discussion is most emphatically NOT about compiled code.  It's about
> the process of compiling sources.

Warnings do not affect compatibility, so no problem there (though I
admit they are ugly).

> What use is this warning message?  How could it prompt a hacker to
> improve his code?  (That's a genuine question, not a rhetorical one.)

It's a warning about non-side-effects code whose result is discarded,
so it obviously prompts the programmer to either check that it didn't
make a mistake (like forgetting to assign the expression's result to a
variable), or remove the code altogether if it is leftover code from a
cut&paste or refactoring.

> What do you mean by "generic" here?  Is the same trick performed on
> symbols other than 'xemacs?

The warning has *nothing* to do with the xemacs symbol. It would still
be there if you did "(and (not (featurep 'cc-fix)) nil ...)". It just
happens to be triggered, in this particular case, by the byte-compiler
optimizing (featurep 'xemacs) to nil. So the warning is entirely
generic.

> Surely it is a bug that "(featurep 'cc-fix)" appears in the message
> rather than "(featurep 'xemacs)".  This situation is not about 'cc-fix,
> it's about 'xemacs.

Absolutely no, as shown above. At the point the byte-compiler throws
the warning, it does not know (I think) where the relevant nil came
from. It is rightly complaining that (featurep 'cc-fix) does nothing
and returns a value that goes unused. It would be possible to store
the original code before optimization and pass it to the warning
generation code, but it is likely not worth it.

> Does
> anybody actually care about "(featurep 'xemacs)" being optimised away?

IIRC, optimizing (featurep 'xemacs) => nil is done, *precisely*, to
help compiling portable code. Because you can have

 (if (featurep 'xemacs)
     ;; lots of code which would throw warnings or errors on Emacs
     ;; because of incompatible parameter profiles and such
   ;; else
   ;; code for emacs

and compile it without getting warnings of errors in code that will
never be executed on Emacs anyway.

    Juanma

bug#6740: Spurious byte compiler warnings

[Alan Mackenzie]

Hi, Juanma,

On Wed, Jul 28, 2010 at 07:56:32PM +0200, Juanma Barranquero wrote:
> > What use is this warning message?  How could it prompt a hacker to
> > improve his code?  (That's a genuine question, not a rhetorical one.)

> It's a warning about non-side-effects code whose result is discarded,
> so it obviously prompts the programmer to either check that it didn't
> make a mistake (like forgetting to assign the expression's result to a
> variable), or remove the code altogether if it is leftover code from a
> cut&paste or refactoring.

OK, thanks.

> > What do you mean by "generic" here?  Is the same trick performed on
> > symbols other than 'xemacs?

> The warning has *nothing* to do with the xemacs symbol. It would still
> be there if you did "(and (not (featurep 'cc-fix)) nil ...)". It just
> happens to be triggered, in this particular case, by the byte-compiler
> optimizing (featurep 'xemacs) to nil. So the warning is entirely
> generic.

I'm not any more doubting the correctness of the message; I'm doubting
its adequacy.  Without understanding that (featurep 'xemacs) has been
optimised to nil, it's impossible to understand the current message
(either of them).  I've spent a long, long time puzzling over this error
message.  If only there were a warning about 'xemacs, it would be plain
and obvious.

> > Does anybody actually care about "(featurep 'xemacs)" being optimised
> > away?

> IIRC, optimizing (featurep 'xemacs) => nil is done, *precisely*, to
> help compiling portable code.

Misunderstanding, sorry!  I'm all in favour of this optimisation being
done.  But I don't care about it, in the sense I don't want to have to
make sense of a confusing message about it.  Does anybody care about
it enough to want that message in this particular case?  Couldn't the
optimisation just be done quietly in the background, with no warning?
Or couldn't there be a warning like

    "`(featurep 'xemacs)' has been translated to nil"

?

> Because you can have

>  (if (featurep 'xemacs)
>      ;; lots of code which would throw warnings or errors on Emacs
>      ;; because of incompatible parameter profiles and such
>    ;; else
>    ;; code for emacs

> and compile it without getting warnings of errors in code that will
> never be executed on Emacs anyway.

OK, I can see that.

>     Juanma

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#6740: Spurious byte compiler warnings

[Juanma Barranquero]

On Wed, Jul 28, 2010 at 21:45, Alan Mackenzie <acm@muc.de> wrote:

> I'm doubting
> its adequacy.  Without understanding that (featurep 'xemacs) has been
> optimised to nil, it's impossible to understand the current message

I think the message is a good hint that something is being statically
determined to be nil inside an `and'.

> (either of them).

Yes, that's a bug :-)

> If only there were a warning about 'xemacs, it would be plain
> and obvious.

But, as I've explained, there cannot (easily) be a waning about
`xemacs'; it would have to be about any code that statically evaluates
to nil in such a context. I'm not sure how clean that would be to
implement, and anyway no one has been bothered enough to try it.

> Does anybody care about
> it enough to want that message in this particular case?

Yes. You don't know whether a warning is relevant or not unless you
get it. In *this* particular case, all you need to quiet the
byte-compiler is

@@ -1,5 +1,5 @@
 (eval-when-compile
-  (if (and (not (featurep 'cc-fix))
-          (featurep 'xemacs)
+  (if (and (featurep 'xemacs)
+          (not (featurep 'cc-fix))
            (progn
              (require 'font-lock)


> Couldn't the
> optimisation just be done quietly in the background, with no warning?

Why? The optimization is detecting something suspicious, and acting
accordingly.

> Or couldn't there be a warning like
>
>    "`(featurep 'xemacs)' has been translated to nil"

????

That's worse that what you're complaining now; every use of (featurep
'xemacs) in the sources would produce warnings!

    Juanma

bug#6740: Spurious byte compiler warnings

[Dan Nicolaescu]

Juanma Barranquero <lekktu@gmail.com> writes:

> On Wed, Jul 28, 2010 at 21:45, Alan Mackenzie <acm@muc.de> wrote:
>
>>
>>    "`(featurep 'xemacs)' has been translated to nil"
>
> ????
>
> That's worse that what you're complaining now; every use of (featurep
> 'xemacs) in the sources would produce warnings!

Exactly, and it would defeat the purpose of optimizing away 
(featurp 'xemacs) (and (featurep 'emacs): being able to have a single source base that 
works on both platforms, and it can be compiled warning free.

bug#6740: Spurious byte compiler warnings

[Alan Mackenzie]

Hi, Juanma,

On Wed, Jul 28, 2010 at 09:54:04PM +0200, Juanma Barranquero wrote:
> On Wed, Jul 28, 2010 at 21:45, Alan Mackenzie <acm@muc.de> wrote:

> > I'm doubting its adequacy.  Without understanding that (featurep
> > 'xemacs) has been optimised to nil, it's impossible to understand the
> > current message

> I think the message is a good hint that something is being statically
> determined to be nil inside an `and'.

No, sorry it isn't.  It's a good hint that there's a bug in the byte
compiler, the tentative conclusion I came to after trying various things.
If it were such a good hint, I wouldn't have spent several hours of
bafflement trying to figure out what was wrong.  Or is it just me who's
uniquely stupid?  Is there anybody here listening in who's seen this
message and immediately understood it?

Abstractly seen, the warning did not relate to my source code; it related
to an internal, transformed, different piece of source created by the
compiler.  I think warning messages should always be wrt the original
code.


> > If only there were a warning about 'xemacs, it would be plain and
> > obvious.

> But, as I've explained, there cannot (easily) be a warning about
> `xemacs'; it would have to be about any code that statically evaluates
> to nil in such a context.

No.  It would be about code which the _compiler_ had transformed to a
static nil.  The cause of my confusion was that silent change to 'xemacs.
There are surely not too many situations where the compiler does this,
are there?

> I'm not sure how clean that would be to implement, and anyway no one
> has been bothered enough to try it.

:-)  I know nothing of the byte compiler, but maybe I'll give it a go
sometime.

> > Does anybody care about it enough to want that message in this
> > particular case?

> Yes. You don't know whether a warning is relevant or not unless you
> get it. In *this* particular case, all you need to quiet the
> byte-compiler is

> @@ -1,5 +1,5 @@
>  (eval-when-compile
> -  (if (and (not (featurep 'cc-fix))
> -          (featurep 'xemacs)
> +  (if (and (featurep 'xemacs)
> +          (not (featurep 'cc-fix))
>             (progn
>               (require 'font-lock)

Yes; I've already made that change, thanks!  But what is the process by
which I'm meant to come to sufficient understanding to be able figure
this out?

> > Couldn't the optimisation just be done quietly in the background,
> > with no warning?

> Why? The optimization is detecting something suspicious, and acting
> accordingly.

There's nothing suspicious about having (featurep 'xemacs) in the middle
of an `and' form.  I agree there would be something suspicious about
having a bare `nil' there.  Can't the compiler figure out the difference
between these two cases somehow?

> > Or couldn't there be a warning like

> >    "`(featurep 'xemacs)' has been translated to nil"

> ????

> That's worse that what you're complaining now; every use of (featurep
> 'xemacs) in the sources would produce warnings!

Oh, all right then.  Can I ask you to come up with some form of warning
which, several years ago, would have saved me all these hours of
frustration?


>     Juanma

-- 
Alan Mackenzie (Nuremberg, Germany).

bug#6740: Spurious byte compiler warnings

[Juanma Barranquero]

On Thu, Jul 29, 2010 at 22:24, Alan Mackenzie <acm@muc.de> wrote:

> No, sorry it isn't.  It's a good hint that there's a bug in the byte
> compiler, the tentative conclusion I came to after trying various things.

No, of course not, because there's no bug (well, other than the
duplicate message).

The warning is obscure, yes. Warnings about optimized code often are.
I agree that it'd be good to have better warnings and error messages
(something like "optimization X has rendered code Y unused", perhaps).
IIRC, there's published research about good error messages in the face
of program transformations.

> If it were such a good hint, I wouldn't have spent several hours of
> bafflement trying to figure out what was wrong.  Or is it just me who's
> uniquely stupid?

It's not about being stupid, only about knowing a bit what the
bytecompiler does, or not. You didn't, and you're *right* that the
warning is difficult to interpret.

> Is there anybody here listening in who's seen this
> message and immediately understood it?

Well, I did, but I remembered the (featurep 'xemacs) optimization so I
had a head start.

> Abstractly seen, the warning did not relate to my source code; it related
> to an internal, transformed, different piece of source created by the
> compiler.  I think warning messages should always be wrt the original
> code.

Agreed.

> The cause of my confusion was that silent change to 'xemacs.
> There are surely not too many situations where the compiler does this,
> are there?

Even if there aren't many, optimizations could be added later; for
example, to statically determine that (= 0 0) is t, or even that (and
(> x 0) (< x 0)) is nil.

> Yes; I've already made that change, thanks!  But what is the process by
> which I'm meant to come to sufficient understanding to be able figure
> this out?

That's a good question.

> There's nothing suspicious about having (featurep 'xemacs) in the middle
> of an `and' form.

No, but the warning isn't about that. Is about unused side-effect-free code.

> I agree there would be something suspicious about
> having a bare `nil' there.  Can't the compiler figure out the difference
> between these two cases somehow?

Yes, it could. Again, no one has bothered to implement it, because
these kinds of problems aren't common (most warnings are relatively
understandable withouth delving into the bytecompiler sources).

> Oh, all right then.  Can I ask you to come up with some form of warning
> which, several years ago, would have saved me all these hours of
> frustration?

Not me. I understand what the bytecompiler is doing (at least, I think
I do) but I'll leave fixing it for the experts.

    Juanma

bug#6740: Spurious byte compiler warnings

[Stefan Monnier]

>> Warning: value returned from (featurep (quote cc-fix)) is unused
>> Warning: value returned from (featurep (quote cc-fix)) is unused
>> 
>> It is obvious that that value is indeed used.  This is a bug.

> The byte compiler knows that (featurep 'xemacs) is false, so 
>  (and (not (featurep 'cc-fix)) ... )
> will be false, so the featurep result is indeed unused.

> If you use (and (featurep 'xemacs) (not (featurep 'cc-fix) ...
> the warning will go away.

It's one of those cases where the warning is the result of a check done
"too late" (i.e. after some optimization), which means the check is not
performed on the code the user sees, but on some massaged version of it.
It's difficult to avoid them, short of removing all "code
improvement" warnings.


        Stefan

bug#6740: Spurious byte compiler warnings

[Stefan Kangas]

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

>>> Warning: value returned from (featurep (quote cc-fix)) is unused
>>> Warning: value returned from (featurep (quote cc-fix)) is unused
>>>
>>> It is obvious that that value is indeed used.  This is a bug.
>
>> The byte compiler knows that (featurep 'xemacs) is false, so
>>  (and (not (featurep 'cc-fix)) ... )
>> will be false, so the featurep result is indeed unused.
>
>> If you use (and (featurep 'xemacs) (not (featurep 'cc-fix) ...
>> the warning will go away.
>
> It's one of those cases where the warning is the result of a check done
> "too late" (i.e. after some optimization), which means the check is not
> performed on the code the user sees, but on some massaged version of it.
> It's difficult to avoid them, short of removing all "code
> improvement" warnings.

(That was 10 years ago.)

I think this is a very minor issue that is likely not worth spending any
time on.  In any case, it's not even clear what the fix would be or if
a different behavior would improve the situation.

Does anyone object to closing this as wontfix?

bug#6740: Spurious byte compiler warnings

[Lars Ingebrigtsen]

Stefan Kangas <stefan@marxist.se> writes:

> I think this is a very minor issue that is likely not worth spending any
> time on.  In any case, it's not even clear what the fix would be or if
> a different behavior would improve the situation.
>
> Does anyone object to closing this as wontfix?

Well, it is a real bug (and somewhat annoying), so I think it should be
left open.

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