bug(s) in moving messages.

bug(s) in moving messages.

[Vladimir Volovich]

Hi,

I just tried the following:

* i wrote a message to myself with russian test
* i received it (into nnmbox:mail.misc). Russian letters looked like
  \214<something>, as i wrote in a prev.letter
* I moved a message to another nnmbox group with B m

[btw, is it a bug in gnus, that after moving a message it is invisible
(inaccessible) until i restart gnus?]

* i restarted gnus to see my moved message (see above)

* the russian text in the _BODY_ of the message was in encoded-word
  encoding:

--------
From:
To:

this is a test
test =?iso-8859-5?b?3LHnILHjsLE=?= this is a test
etc
--------

This is definitely a bug.

	Best regards, -- Vladimir.
-- 
Without followers, evil cannot spread.
		-- Spock, "And The Children Shall Lead", stardate 5029.5

Re: bug(s) in moving messages.

[Lars Magne Ingebrigtsen]

Vladimir Volovich <vvv@vvv.vsu.ru> writes:

> * i wrote a message to myself with russian test
> * i received it (into nnmbox:mail.misc). Russian letters looked like
>   \214<something>, as i wrote in a prev.letter

Could you look in the `nnmbox-mbox-buffer' buffer and see how the
Russian characters look there?

> * the russian text in the _BODY_ of the message was in encoded-word
>   encoding:

I think this has been fixed in 0.28, but could you check?

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

Re: bug(s) in moving messages.

[Vladimir Volovich]

"LMI" == Lars Magne Ingebrigtsen writes:

 LMI> Vladimir Volovich <vvv@vvv.vsu.ru> writes:
 >> * i wrote a message to myself with russian test * i received it
 >> (into nnmbox:mail.misc). Russian letters looked like
 >> \214<something>, as i wrote in a prev.letter

 LMI> Could you look in the `nnmbox-mbox-buffer' buffer and see how
 LMI> the Russian characters look there?

I visited the mbox buffer, and my message there looked like:

--------------------------------------------------------
>From vvv  Sat Sep 12 13:32:01 1998
Return-Path: <vvv>
Received: (from vvv@localhost)
	by vvv.vsu.ru (8.8.7/8.8.7) id NAA02604;
	Sat, 12 Sep 1998 13:32:01 +0400
To: vvv
Subject: =?iso-8859-5?b?4tXh4g==?= test =?iso-8859-5?b?3+De0dAg39Xg0A==?=
From: Vladimir Volovich <vvv@vvv.vsu.ru>
Date: 12 Sep 1998 13:32:00 +0400
Message-ID: <m3n285fzhr.fsf@vvv.vsu.ru>
User-Agent: Gnus/5.070028 (Pterodactyl Gnus v0.28) Emacs/20.3
MIME-Version: 1.0
Content-Type: text/plain; charset="iso-8859-5"
Content-Transfer-Encoding: 8bit
Lines: 8
Xref: vvv.vsu.ru mail.misc:2064
X-Gnus-Newsgroup: mail.misc:2064   Sat Sep 12 13:32:02 1998

Hi,

мБч БуАБ.  Ючяп ъуЮп.  ЮчруЮзп АрОвь.

	Best regards, -- Vladimir.
------------------------------------------------------------------

I.e. the text was 8-bit encoded with iso-8859-5. However, the above is
a pasted text from the mbox buffer. When i was visiting mbox, i saw
instead of russian characters combos like

\315\342\336 \342\325\341\342. ...

i.e. each character's octal code (i did not touch standard-display-8bit).

In my nnmbox:mail.misc russian text in this message looked like

\214<8-bit_char>\214<8-bit_char>\214<8-bit_char> \214<8-bit_char>...

However, in a summary buffer of nnmbox:mail.misc, the subject in
russian looked Ok. The *same* message in nnml:archive looked
completely correct!

 >> * the russian text in the _BODY_ of the message was in
 >> encoded-word encoding:

 LMI> I think this has been fixed in 0.28, but could you check?

Well, the behavior have changed. Now gnus does not use encode-word for
message body, but i could not check whether it moves correctly because
of those nasty \214<8-bit_char> representation in my nnmbox backend...

For the first look, it seems that some russian characters disappear
while moving. I'll answer when i'll be able to read my articles. :-)

	Best regards, -- Vladimir.

Re: bug(s) in moving messages.

[Lars Magne Ingebrigtsen]

Vladimir Volovich <vvv@vvv.vsu.ru> writes:

> I visited the mbox buffer, and my message there looked like:

[...]

> i.e. each character's octal code (i did not touch standard-display-8bit).

That's how it should be there, then...

> In my nnmbox:mail.misc russian text in this message looked like
> 
> \214<8-bit_char>\214<8-bit_char>\214<8-bit_char> \214<8-bit_char>...

But not there, so it would seem that something goes wrong while
copying from the mbox buffer to the article buffer somewhere.  It
looks like the text has been decoded twice.

Could you look in the " *nntpd*" (note leading space) buffer after
selecting one of these articles to see how it looks there?

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

Re: bug(s) in moving messages.

[Vladimir Volovich]

"LMI" == Lars Magne Ingebrigtsen writes:

 >> In my nnmbox:mail.misc russian text in this message looked like
 >> 
 >> \214<8-bit_char>\214<8-bit_char>\214<8-bit_char>
 >> \214<8-bit_char>...

 LMI> But not there, so it would seem that something goes wrong while
 LMI> copying from the mbox buffer to the article buffer somewhere.
 LMI> It looks like the text has been decoded twice.

 LMI> Could you look in the " *nntpd*" (note leading space) buffer
 LMI> after selecting one of these articles to see how it looks there?

It appeared there without a message body:

[some stuff skipped]
221 2072 Article retrieved.
>From vvv  Sat Sep 12 17:23:37 1998
Return-Path: <vvv>
Received: (from vvv@localhost) by vvv.vsu.ru (8.8.7/8.8.7) id RAA09133; Sat, 12 Sep 1998 17:23:37 +0400
To: vvv
Subject: test =?iso-8859-5?b?4tXh4g==?=
From: Vladimir Volovich <vvv@vvv.vsu.ru>
Date: 12 Sep 1998 17:23:37 +0400
Message-ID: <m33e9xh3c6.fsf@vvv.vsu.ru>
User-Agent: Gnus/5.070028 (Pterodactyl Gnus v0.28) Emacs/20.3
MIME-Version: 1.0
Content-Type: text/plain; charset="iso-8859-5"
Content-Transfer-Encoding: 8bit
Lines: 7
Xref: vvv.vsu.ru mail.misc:2072
X-Gnus-Newsgroup: mail.misc:2072   Sat Sep 12 17:23:38 1998
.
[end of buffer]

BTW, this message i receive not via nntpd, but from my mail spool
dir. Should the " *nntpd*" buffer contain info about non-nntp
messages?

	Best regards, -- Vladimir.

Re: bug(s) in moving messages.

[Lars Magne Ingebrigtsen]

Vladimir Volovich <vvv@vvv.vsu.ru> writes:

>  LMI> Could you look in the " *nntpd*" (note leading space) buffer
>  LMI> after selecting one of these articles to see how it looks there?
> 
> It appeared there without a message body:

Right; it copies the things directly into the article buffer.  Could
you look in the " *Original Article whatever*" buffer instead?

> BTW, this message i receive not via nntpd, but from my mail spool
> dir. Should the " *nntpd*" buffer contain info about non-nntp
> messages?

The " *nntpd*" buffer is used for all communication between Gnus and
its backends.

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

Re: bug(s) in moving messages.

[Владимир Волович]

"LMI" == Lars Magne Ingebrigtsen writes:

 LMI> Right; it copies the things directly into the article buffer.
 LMI> Could you look in the " *Original Article whatever*" buffer
 LMI> instead?

There was only a buffer " *Original Article*" (without whatever :).
The situation is following:

For those articles which look correctly (e.g. in nntp and nnml
backends), russian letters in the " *Original Article*" buffer look
like "\323\320\301\323..." i.e. like unencoded 8-bit characters
characters. In the Article buffer they become russian letter (after
decoding).

But for the nnmbox backend, russian letters in the " *Original
Article*" buffer look already decoded, i.e. like perfect russian
letters! But in the article buffer they instead look like a mess with
\214<8-bit_char>. So yes, it seems that the error is that the article
gets decoded twice: the first time in the " *Original Article*"
buffer, and the second time in the Article buffer. :-(

	Best regards, -- Vladimir.

Re: bug(s) in moving messages.

[Lars Magne Ingebrigtsen]

"Владимир Волович" <vvv@vvv.vsu.ru> writes:

> But for the nnmbox backend, russian letters in the " *Original
> Article*" buffer look already decoded, i.e. like perfect russian
> letters! But in the article buffer they instead look like a mess with
> \214<8-bit_char>. So yes, it seems that the error is that the article
> gets decoded twice: the first time in the " *Original Article*"
> buffer, and the second time in the Article buffer. :-(

I think I've located and fixed this bug.  Could you check after
Pterodactyl Gnus v0.31 has been released to see whether things start
working?

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

Re: bug(s) in moving messages.

[Владимир Волович]

"LMI" == Lars Magne Ingebrigtsen writes:

 LMI> " ŒшŒпŒтŒьŒэŒьŒЮ
 LMI> ŒчŒшŒчŒрŒьŒГ" <vvv@vvv.vsu.ru> writes:
 >> But for the nnmbox backend, russian letters in the " *Original
 >> Article*" buffer look already decoded, i.e. like perfect russian
 >> letters! But in the article buffer they instead look like a mess
 >> with \214<8-bit_char>. So yes, it seems that the error is that the
 >> article gets decoded twice: the first time in the " *Original
 >> Article*" buffer, and the second time in the Article buffer. :-(

 LMI> I think I've located and fixed this bug.  Could you check after
 LMI> Pterodactyl Gnus v0.31 has been released to see whether things
 LMI> start working?

Unfortunately, no. Things seem to work like previously (including
buffers " *Original Article*"). Some more details follow.

1) Some messages in russian display in nnmbox groups well. For
example, the message with

MIME-Version: 1.0
Content-Type: multipart/alternative;
	boundary="----=_NextPart_000_0010_01BDDFDF.5FBF1020"

and the text part was

Content-Type: text/plain;
	charset="koi8-r"
Content-Transfer-Encoding: quoted-printable

This message displayed perfectly well. (and i suppose that all qp
messages will be ok). problems are with 8-bit messages.

2) subjects and "from" addresses of the messages which are in
   encoded-word format, are displayed in a summary buffer and in an
   article buffer; the _body_ is still displayed broken.

3) if the message is text/plain charset=koi8-r, and c-t-e=8-bit, then
   these messages are looking well in an " *Original Article*" buffer
   (i.e., russian letters are looking Ok).

4) if the message is text/plain charset=iso-8859-5, and c-t-e=8-bit,
   then russian letters in an " *Original Article*" buffer look like
   russian letters, but encoded brokenly (as if they were iso-8859-5
   text displayed on a koi-r terminal).

my setup:

(set-default-font "fontset-standard")
(set-language-environment "Cyrillic-KOI8")
(setq rfc2047-default-charset 'koi8-r)
;(setq mm-default-coding-system 'koi8) ; this problem is independent of this setting
(setq gnus-secondary-select-methods '((nnmbox "")))

	Best regards, -- Vladimir.