Percent sign in message

Percent sign in message

[Whitfield Diffie]

				Sunday  10 September 2006  at 23:03

    I am sending this message to emacs-devel rather than bug-gnu-emacs
because, although I think this is arguable a bug, it could be fixed by
changing the documentation but the fix I propose is more substantial.

    The command
			       (message "%")

produces the error

	     (error "Not enough arguments for format string")

because message hands its arguments directly to format and format makes
special use of percent signs.  A reasonably careful look at the
documentation of format finds the warning that if you want a percent
in the format string, you should double it.  That isn't so bad because
there is no reason to use format if you don't want the special way it
processes its string.

    Message is another thing.  It is perfectly reasonable to call message
to print a fixed string or, as produced the trouble, a string that is not
under the program's control.  In my case, I have a function called
`describe-last-kbd-macro' that constructs a definition and calls message to
present it and my macro operated on comments in TeX.  Clearly I could
pre-process the message to double the percent signs but it doesn't seem
that should be necessary.

    What I propose is changing format so that it specially interprets as
many percent signs as it has arguments for and just passes the rest
verbatum, with a possible warning in the *Messages* buffer.

					Whit

Re: Percent sign in message

[Michaël Cadilhac]

[-- Attachment #1.1: Type: text/plain, Size: 719 bytes --]

Whitfield Diffie <whitfield.diffie@sun.com> writes:

> In my case, I have a function called `describe-last-kbd-macro' that
> constructs a definition and calls message to present it and my macro
> operated on comments in TeX. Clearly I could pre-process the message
> to double the percent signs but it doesn't seem that should be
> necessary.

Why don't you use (message "%s" whatever)?

-- 
 |      Michaël `Micha' Cadilhac   |  Mieux vaut se taire                   |
 |         Epita/LRDE Promo 2007   |   Que de parler trop fort.             |
 |  http://michael.cadilhac.name   |           -- As de trèfle              |
 `--  -   JID: micha@amessage.be --'                                   -  --'

[-- Attachment #1.2: Type: application/pgp-signature, Size: 188 bytes --]

[-- Attachment #2: Type: text/plain, Size: 142 bytes --]

_______________________________________________
Emacs-devel mailing list
Emacs-devel@gnu.org
http://lists.gnu.org/mailman/listinfo/emacs-devel

Re: Percent sign in message

[Whitfield Diffie]

				Sunday  10 September 2006  at 23:55

> Why don't you use (message "%s" whatever)?

    Because I'm an idiot.  Thank you.

					Whit

Re: Percent sign in message

[David Kastrup]

Whitfield Diffie <whitfield.diffie@sun.com> writes:

> 				Sunday  10 September 2006  at 23:55
>
>> Why don't you use (message "%s" whatever)?
>
>     Because I'm an idiot.  Thank you.

Well, we have had enough "idiots" asking this question by now that it
would not seem amiss to mention this idiom in DOC-string and/or Elisp
manual entry.  It does not seem immediately obvious.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: Percent sign in message

[Kim F. Storm]

David Kastrup <dak@gnu.org> writes:

>>> Why don't you use (message "%s" whatever)?
>>
> Well, we have had enough "idiots" asking this question by now that it
> would not seem amiss to mention this idiom in DOC-string and/or Elisp
> manual entry.  It does not seem immediately obvious.

Done.

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

Re: Percent sign in message

[Stefan Monnier]

>>>> Why don't you use (message "%s" whatever)?
>> Well, we have had enough "idiots" asking this question by now that it
>> would not seem amiss to mention this idiom in DOC-string and/or Elisp
>> manual entry.  It does not seem immediately obvious.

Of course, we could also add a compiler warning whenever we see
(message <exp>) and <exp> is not a constant string.


        Stefan

Re: Percent sign in message

[Richard Stallman]

    Of course, we could also add a compiler warning whenever we see
    (message <exp>) and <exp> is not a constant string.

I think that would give too many false warnings, because it is not
unusual to obtain the format string from some non-constant source.

Re: Percent sign in message

[Stefan Monnier]

>     Of course, we could also add a compiler warning whenever we see
>     (message <exp>) and <exp> is not a constant string.

> I think that would give too many false warnings, because it is not
> unusual to obtain the format string from some non-constant source.

Could be.  But the warning can be turned off by using (message "%s" <exp>).


        Stefan

Re: Percent sign in message

[David Kastrup]

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

>>     Of course, we could also add a compiler warning whenever we see
>>     (message <exp>) and <exp> is not a constant string.
>
>> I think that would give too many false warnings, because it is not
>> unusual to obtain the format string from some non-constant source.
>
> Could be.  But the warning can be turned off by using (message "%s" <exp>).

That would need to be
(message "%s" (format <exp>))

It is not unusual to obtain a _format_ string from some non-constant
source.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: Percent sign in message

[Stefan Monnier]

>>> Of course, we could also add a compiler warning whenever we see
>>> (message <exp>) and <exp> is not a constant string.
>> 
>>> I think that would give too many false warnings, because it is not
>>> unusual to obtain the format string from some non-constant source.
>> 
>> Could be.  But the warning can be turned off by using (message "%s" <exp>).

> That would need to be
> (message "%s" (format <exp>))

Well, only if <exp> contains escaped % signs.  It's generally not the case.

> It is not unusual to obtain a _format_ string from some non-constant
> source.

But it then comes with some arguments as well, doesn't it?


        Stefan

Re: Percent sign in message

[David Kastrup]

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

>>>> Of course, we could also add a compiler warning whenever we see
>>>> (message <exp>) and <exp> is not a constant string.
>>> 
>>>> I think that would give too many false warnings, because it is not
>>>> unusual to obtain the format string from some non-constant source.
>>> 
>>> Could be.  But the warning can be turned off by using (message "%s" <exp>).
>
>> That would need to be
>> (message "%s" (format <exp>))
>
> Well, only if <exp> contains escaped % signs.  It's generally not
> the case.
>
>> It is not unusual to obtain a _format_ string from some non-constant
>> source.
>
> But it then comes with some arguments as well, doesn't it?

So what you mean is not
  "the warning can be turned off by using (message "%s" <exp>)"
but rather
  "the warning can be turned off by using (message <exp> <exp>+)"

Namely: warn if we use a _single_ non-constant expression as argument,
but don't complain if we use more than one expression as argument.

I'd consider that a good idea, but I find it hard to gather this as
your proposal from the wording you used.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: Percent sign in message

[Stefan Monnier]

> So what you mean is not
>   "the warning can be turned off by using (message "%s" <exp>)"
> but rather
>   "the warning can be turned off by using (message <exp> <exp>+)"

No I really meant just what I said:

  Output a warning when we see (message <exp>).

And such a warning can be silenced by replacing the offending expression
with (message "%s" <exp>).

> Namely: warn if we use a _single_ non-constant expression as argument,

Isn't that what I said:
 
  whenever we see (message <exp>) and <exp> is not a constant string


-- Stefan

Re: Percent sign in message

[Richard Stallman]

    > I think that would give too many false warnings, because it is not
    > unusual to obtain the format string from some non-constant source.

    Could be.  But the warning can be turned off by using (message "%s" <exp>).

Hmm.  If the warning is given only in the case of one arg, maybe it won't
get too many spurious hits.  I was thinking of getting nontrivial format
strings from nonconstant sources; but those cases would usually have
further arguments.

But let's save this for later.

Re: Percent sign in message

[David Kastrup]

Richard Stallman <rms@gnu.org> writes:

>     > I think that would give too many false warnings, because it is not
>     > unusual to obtain the format string from some non-constant source.
>
>     Could be.  But the warning can be turned off by using (message
>     "%s" <exp>).
>
> Hmm.  If the warning is given only in the case of one arg, maybe it won't
> get too many spurious hits.  I was thinking of getting nontrivial format
> strings from nonconstant sources; but those cases would usually have
> further arguments.
>
> But let's save this for later.

I'd want to mention that format string vulnerabilities are a common
attack vector for viruses.  They will not likely be exploitable in
Elisp, but it shows that they are not rare among programmers (indeed,
one such case prompted the release of Emacs 21.4), and could cause
weird effects.

I think such a warning would not be amiss, and should probably be
extended to `error' as well.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

Re: Percent sign in message

[Stuart D. Herring]

> I'd want to mention that format string vulnerabilities are a common
> attack vector for viruses.  They will not likely be exploitable in
> Elisp, but it shows that they are not rare among programmers (indeed,
> one such case prompted the release of Emacs 21.4), and could cause
> weird effects.

This is off-topic, but what was the case that prompted 21.4?  I've never
seen it described.

Thanks,
Davis

-- 
This product is sold by volume, not by mass.  If it appears too dense or
too sparse, it is because mass-energy conversion has occurred during
shipping.

Re: Percent sign in message

[David Kastrup]

"Stuart D. Herring" <herring@lanl.gov> writes:

>> I'd want to mention that format string vulnerabilities are a common
>> attack vector for viruses.  They will not likely be exploitable in
>> Elisp, but it shows that they are not rare among programmers (indeed,
>> one such case prompted the release of Emacs 21.4), and could cause
>> weird effects.
>
> This is off-topic, but what was the case that prompted 21.4?  I've never
> seen it described.

Well, the diff is short enough.  A format string vulnerability in the
movemail executable.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum