bug#7256: Convert non-ASCII letters to hexadecimal? vs control chars.

bug#7256: Convert non-ASCII letters to hexadecimal? vs control chars.

[jidanni]

>>>>> "LMI" == Lars Magne Ingebrigtsen <larsi@gnus.org> writes:
LMI> jidanni@jidanni.org writes:

>> "Convert non-ASCII letters to hexadecimal? (y or n)  y" will still pass
>> control characters. Yes they are ASCII, but who usually wants to pass them?

LMI> grep -nH -e "Convert non-ASCII" *.el

LMI> Grep finished with no matches found at Tue Oct 19 18:45:12

24.0.50/lisp/mail/emacsbug.el.gz:345:              (y-or-n-p "Convert non-ASCII letters to hexadecimal? "))

    ;; Check the buffer contents and reject non-English letters.
    ;; FIXME message-mode probably does this anyway.
    (goto-char (point-min))
    (skip-chars-forward "\0-\177")
    (unless (eobp)
      (if (or report-emacs-bug-no-confirmation
              (y-or-n-p "Convert non-ASCII letters to hexadecimal? "))
          (while (progn (skip-chars-forward "\0-\177")
                        (not (eobp)))
            (let ((ch (following-char)))
              (delete-char 1)
              (insert (format "=%02x" ch))))))

Anyway, the typical scenario is the user is asked to submit a backtrace,
and of course backtraces are full of nasty (useful, even?) binary mess
characters, which triggers multiple question upon trying to send, some
from the above, some from message.el.

Anyway, looks like time to fix the FIXME above.

Re: Convert non-ASCII letters to hexadecimal? vs control chars.

[Lars Magne Ingebrigtsen]

jidanni@jidanni.org writes:

>       (if (or report-emacs-bug-no-confirmation
>               (y-or-n-p "Convert non-ASCII letters to hexadecimal? "))

[...]

> Anyway, the typical scenario is the user is asked to submit a backtrace,
> and of course backtraces are full of nasty (useful, even?) binary mess
> characters, which triggers multiple question upon trying to send, some
> from the above, some from message.el.

Yeah, that seems really suboptimal, but I'm not sure what the fix should
be.  Converting everything to hex seems rather odd, since that would
make látin and unicode characters also be hexified, which is weird.

I think the right fix is to remove this question altogether from
emacsbug.el, and inhibit the "remove non-printable characters" query
from message.el.

Any objections?

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

bug#7256: Convert non-ASCII letters to hexadecimal? vs control chars.

[Eli Zaretskii]

> From: Lars Magne Ingebrigtsen <larsi@gnus.org>
> Date: Thu, 21 Oct 2010 18:14:21 +0200
> Cc: bug-gnu-emacs@gnu.org, rfrancoise@debian.org, bugs@gnus.org,
> 	emacs-devel@gnu.org
> 
> Yeah, that seems really suboptimal, but I'm not sure what the fix should
> be.  Converting everything to hex seems rather odd, since that would
> make látin and unicode characters also be hexified, which is weird.

I think the right fix for control characters below 32 is to convert
them to the ^X ASCII representation.

Re: Convert non-ASCII letters to hexadecimal? vs control chars.

[Stefan Monnier]

> I think the right fix for control characters below 32 is to convert
> them to the ^X ASCII representation.

I'm not sure I'd agree enough to make it a promptless default.


        Stefan

Re: Convert non-ASCII letters to hexadecimal? vs control chars.

[Lars Magne Ingebrigtsen]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:

>> I think the right fix for control characters below 32 is to convert
>> them to the ^X ASCII representation.
>
> I'm not sure I'd agree enough to make it a promptless default.

I think it might make sense for the bug reports, but I think that (in
general) there's no point in doing any conversion of anything.  I mean,
the mail will be encoded with a proper transfer encoding when sending,
won't it?

Unless one of the sending methods used by emacsbugs doesn't encode
anything. 

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: Convert non-ASCII letters to hexadecimal? vs control chars.

[Stefan Monnier]

>>> I think the right fix for control characters below 32 is to convert
>>> them to the ^X ASCII representation.
>> I'm not sure I'd agree enough to make it a promptless default.
> I think it might make sense for the bug reports, but I think that (in
> general) there's no point in doing any conversion of anything.  I mean,
> the mail will be encoded with a proper transfer encoding when sending,
> won't it?

Yes.  And if not, then it's a separate problem.


        Stefan

Re: Convert non-ASCII letters to hexadecimal? vs control chars.

[Lars Magne Ingebrigtsen]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:

> Yes.  And if not, then it's a separate problem.

So you're fine with just removing the "encode in hex" question?

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen

Re: Convert non-ASCII letters to hexadecimal? vs control chars.

[Stefan Monnier]

>> Yes.  And if not, then it's a separate problem.
> So you're fine with just removing the "encode in hex" question?

Not sure I understand, but I'm fine with sending random bytes, without
removing them first and without asking the user, yes.  We can either
assume they'll get QP or base64 encoded down the line or that it doesn't
matter.


        Stefan

Re: Convert non-ASCII letters to hexadecimal? vs control chars.

[Eli Zaretskii]

> From: Stefan Monnier <monnier@iro.umontreal.ca>
> Date: Thu, 21 Oct 2010 23:22:24 -0400
> 
> I'm fine with sending random bytes, without removing them first and
> without asking the user, yes.  We can either assume they'll get QP
> or base64 encoded down the line or that it doesn't matter.

FWIW, for me a bug report encoded in base64 is a non-starter.  Even qp
is a turn-off.  But that's me.