bug?

bug?

[Vladimir Volovich]

Hi,

something strange happened. i pressed "g" in a group buffer to get new
mail, but Emacs asked "Revisit the file literally?" (what does it mean
and why this happened?). I answered yes. Then the history of my nnmbox
groups was broken. If i try to exit gnus (q), Emacs asks the same
question. I turned on debugging, and when i answered "no" to this
question, an error occurs "File already visited non-literally".
Backtrace follows.

Signaling: (error "File already visited non-literally")
  signal(error ("File already visited non-literally"))
  error("File already visited non-literally")
  find-file-noselect("/home/vvv/mbox" nil raw)
  apply(find-file-noselect ("/home/vvv/mbox" nil raw))
  nnheader-find-file-noselect("/home/vvv/mbox" nil raw)
  nnmbox-read-mbox()
  nnmbox-request-scan(nil "")
  gnus-request-scan(nil (nnmbox ""))
  gnus-read-active-file()
  gnus-group-get-new-news(nil)
* call-interactively(gnus-group-get-new-news)

??? I'm confused... ;-)

	Best regards, -- Vladimir.

bug: coding system for mbox?

[Vladimir Volovich]

Hello,

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

> something strange happened. i pressed "g" in a group buffer to get new
> mail, but Emacs asked "Revisit the file literally?" (what does it mean
> and why this happened?). I answered yes. Then the history of my nnmbox
> groups was broken.

This error happens each time gnus tries to read mail from mbox. In
particular, each time i start gnus. :-( I.e. i start emacs with the
following ~/.emacs file:

---------------------------
(set-default-font "fontset-standard")
(set-language-environment "Cyrillic-KOI8")

(setq nnmail-delete-incoming t)
(setq rfc2047-default-charset 'koi8-r)

(setq gnus-select-method '(nntp "localhost"))
(setq gnus-secondary-select-methods '((nnmbox "")))
---------------------------

There is no ~/.gnus* files. Then i turn on debugging. Then i start
Gnus: M-x gnus. At some point of loading gnus, Emacs asks: "Revisit
the file literally?". If i answer yes, then Emacs `thinks' quite a
while (my mbox file is currently 13Mb, but this is not a problem ---
sometimes i had even bigger ones without problems). After `thinking',
gnus loads normally, and i can more-or-less normally read mail/news
[but there are usually some overestimations of numbers of articles in
my nnmbox groups which gnus reports], until i press `g' in the group
buffer to get new news/mail, in which case the situation repeats
(Emacs again asks that nasty question!).  But if i press `n' on that
question, an error occurs. Here is another backtrace which i made
after fresh start of emacs, then turning on debuggning, starting gnus
and answering `n' to that question:

Signaling: (error "File already visited non-literally")
  signal(error ("File already visited non-literally"))
  error("File already visited non-literally")
  find-file-noselect("/home/vvv/mbox" nil raw)
  apply(find-file-noselect ("/home/vvv/mbox" nil raw))
  nnheader-find-file-noselect("/home/vvv/mbox" nil raw)
  nnmbox-read-mbox()
  nnmbox-request-scan(nil "")
  gnus-request-scan(nil (nnmbox ""))
  gnus-read-active-file(nil nil)
  gnus-setup-news(nil nil nil)
  byte-code("\b„\x0e
  gnus-1(nil nil nil)
  gnus(nil)
  call-interactively(gnus)
  execute-extended-command(nil)
* call-interactively(execute-extended-command)

May be, some contents of mbox makes Emacs to think that mbox is in
some particular coding system (there is some AI?). Maybe, gnus should
force a `literal' opening of mbox? But these are only is my guesses.
If i'll remove mbox and start from the new one, the error will
disappear (but i cannot start from fresh mbox right now; and this bug
should be corrected in any case). *Please* help! This is definitely a
bug of gnus, because it appears in a very clear situation (with that
primitive .emacs file).


 	Best regards, -- Vladimir.

Re: bug: coding system for mbox?

[Vladimir Volovich]

"VV" == Vladimir Volovich writes:

 VV> This error happens each time gnus tries to read mail from
 VV> mbox. In particular, each time i start gnus. :-( I.e. i start
 VV> emacs with the following ~/.emacs file:

One note: this error _is_ definitely a bug in gnus 0.33; it is absent
if i use semi-gnus (pgnus-ichikawa), which is base on exactly this
version of pgnus (i.e. 0.33):

T-gnus 6.10.022 (based on Pterodactyl Gnus 0.33 ; for SEMI 1.8, FLIM 1.8/1.9)

i.e. in pgnus-ichikawa i do not have problems with reading mbox;
changing pgnus-ichikawa to pgnus leads to this error.

	Best regards, -- Vladimir.

semi-pgnus?

[Steinar Bang]

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

> One note: this error _is_ definitely a bug in gnus 0.33; it is absent
> if i use semi-gnus (pgnus-ichikawa), which is base on exactly this
> version of pgnus (i.e. 0.33):

> T-gnus 6.10.022 (based on Pterodactyl Gnus 0.33 ; for SEMI 1.8, FLIM 1.8/1.9)

Is there a semi-gnus based on pgnus?

What does it do? Strip out the native MIME support, and replace it
with SEMI?

Re: semi-pgnus?

[Vladimir Volovich]

"SB" == Steinar Bang writes:

 SB> Is there a semi-gnus based on pgnus?

Yes.

 SB> What does it do? Strip out the native MIME support, and replace
 SB> it with SEMI?

Sort of. However, synchrocization with pgnus because of this is not
perfect: some bugs corrected in pgnus are still bugs in semi-gnus
based on pgnus.

	Best regards, -- Vladimir.

Re: bug: coding system for mbox?

[Lars Magne Ingebrigtsen]

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

> Signaling: (error "File already visited non-literally")

[...]

> May be, some contents of mbox makes Emacs to think that mbox is in
> some particular coding system (there is some AI?). Maybe, gnus should
> force a `literal' opening of mbox? But these are only is my guesses.

>From reading the source, it looks to me like nnmbox is only opening
that file literally.

Could you `M-x debug-on-entry RET find-file-noselect RET' and then
start Gnus?  Look for calls that try to find the mbox file, and see
whether any of these lack the `raw' param.

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

Re: bug: coding system for mbox?

[Vladimir Volovich]

"LMI" == Lars Magne Ingebrigtsen writes:

 LMI>  From reading the source, it looks to me like nnmbox is only
 LMI> opening that file literally.

 LMI> Could you `M-x debug-on-entry RET find-file-noselect RET' and
 LMI> then start Gnus?  Look for calls that try to find the mbox file,
 LMI> and see whether any of these lack the `raw' param.

i've checked this, and find-file-noselect was called three times after
starting gnus; all times with "nil raw" parameters. The third time an
error occurs... [on semi-gnus find-file-noselect gets called only
twice before the group buffer appears].

	Best regards, -- Vladimir.

Re: bug: coding system for mbox?

[Vladimir Volovich]

"LMI" == Lars Magne Ingebrigtsen writes:

 LMI> Could you `M-x debug-on-entry RET find-file-noselect RET' and
 LMI> then start Gnus?  Look for calls that try to find the mbox file,
 LMI> and see whether any of these lack the `raw' param.

Ha! I've found a message which triggers such bug! :-) I'm sending it
to you in a private mail, so you'll be able to test and fix this bug!
If this message is present in a mbox file, gnus immediately starts to
behave like i described (i.e. each time i press `g' in a group buffer
and/or each time i start gnus, a described error occurs); removing it
from mbox (deleting) restores `normal' behavior.

	Best regards, -- Vladimir.