draft group weirdness

draft group weirdness

[Eric Abrahamsen]

I've been having a lot of trouble with the drafts group recently. I
save drafts, go back into the drafts folder, hit 'e' to edit a message,
and it asks me for a file to save into. If I quit out of that with C-g
it doesn't open the message for editing, but makes a copy of the message
buffer in the background. If I mess with it a bit, that can end up
making six or seven copies of the message buffer, all in the background.

If I do choose a file to save into, it appears to save the drafts
summary buffer itself, not the message. I don't know why that would be.

When it happened just now, it also removed all the marks from my
mail.misc group for good measure. All I did was go into drafts, hit 'e',
quit out of the save-to-file prompt, then 'q' to leave my drafts group,
and there were all the messages in mail.misc, fresh as the day I'd
received them!

I'm on Emacs 23 on Ubuntu, running No Gnus. I don't think I have any
drafts-related customization going on. My "nndrafts" server has
"delayed", "drafts" and "queue" in it, that's all.

Any suggestions? Before I lose more marks!

Thanks,
Eric

Re: draft group weirdness

[Richard Riley]

Eric Abrahamsen <eric@ericabrahamsen.net> writes:

> I've been having a lot of trouble with the drafts group recently. I
> save drafts, go back into the drafts folder, hit 'e' to edit a message,
> and it asks me for a file to save into. If I quit out of that with C-g
> it doesn't open the message for editing, but makes a copy of the message
> buffer in the background. If I mess with it a bit, that can end up
> making six or seven copies of the message buffer, all in the background.
>
> If I do choose a file to save into, it appears to save the drafts
> summary buffer itself, not the message. I don't know why that would be.
>

Drafts keeps breaking for me too. I cant pin point it. One thing to
check is that it still in "draft mode". Do a G E from the group buffer
and see if looks like

,----
| ("nndraft:drafts" 1 nil nil
|  (nndraft "")
|  ((timestamp 19866 54941)
|   (gnus-dummy
|    (gnus-draft-mode))))
`----

It might be a start.

More than once my queue has lost its draft mode status - but at the time
I was recreating it when trying to figure out differences between
gnus-select-method and the secondary ones in terms of agent usage.

Re: draft group weirdness

[Lars Magne Ingebrigtsen]

Eric Abrahamsen <eric@ericabrahamsen.net> writes:

> I've been having a lot of trouble with the drafts group recently. I
> save drafts, go back into the drafts folder, hit 'e' to edit a message,
> and it asks me for a file to save into.

That's bizarre.  (setq debug-on-quit t) and `C-g' when it prompts, so
that we can see what function you're calling with `e'.

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

Re: draft group weirdness

[Eric Abrahamsen]

On Wed, Apr 06 2011, Lars Magne Ingebrigtsen wrote:

> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>
>> I've been having a lot of trouble with the drafts group recently. I
>> save drafts, go back into the drafts folder, hit 'e' to edit a message,
>> and it asks me for a file to save into.
>
> That's bizarre.  (setq debug-on-quit t) and `C-g' when it prompts, so
> that we can see what function you're calling with `e'.

Here's the backtrace:

#+begin_src lisp
Debugger entered--Lisp error: (quit)
  ad-Orig-completing-read("File to save in: ~/" (("dummy" . 1)) nil nil nil ido-file-history nil nil)
  completing-read("File to save in: ~/" (("dummy" . 1)) nil nil nil ido-file-history)
  byte-code("\304\305\b	\"\306\307\211\n\307\v&\207" [item prompt ido-text-init history completing-read ido-make-prompt (("dummy" . 1)) nil] 7)
  ido-read-internal(file "File to save in: " ido-file-history nil nil nil)
  ido-read-file-name("File to save in: " "~/" nil nil nil nil)
  read-file-name("File to save in: ")
  basic-save-buffer()
  save-buffer()
  gnus-draft-edit-message()
  call-interactively(gnus-draft-edit-message nil nil)
#+end_src


'G E' on the drafts buffer gives me:


#+begin_src lisp
("nndraft:drafts" 3
 ((1 . 2))
 ((seen 3))
 (nndraft "")
 ((gnus-dummy
   (gnus-draft-mode))))
#+end_src

Hope something in there is a red flag!

Thanks,
Eric

quoting code blocks in Gnus without Org functions (was: draft group weirdness)

[Ted Zlatanov]

On Wed, 06 Apr 2011 09:43:30 +0800 Eric Abrahamsen <eric@ericabrahamsen.net> wrote: 

EA> Here's the backtrace:
EA> #+begin_src lisp
...
EA> #+end_src

I defined this skeleton for myself:

(define-skeleton skeleton-src-quote
  "Insert a quoted Gnus source listing" "language: "
  \n
  > "#+begin_src " str \n
  > "(CODE GOES HERE)" \n
  > "#+end_src" \n \n)

I wonder if there's an easier way, maybe even with language completion
or auto-guessing based on the original buffer's mode, that doesn't
involve Org functions?  IIRC Julien posted how to do it with Org.

Thanks
Ted

Re: quoting code blocks in Gnus without Org functions

[Julien Danjou]

[-- Attachment #1: Type: text/plain, Size: 381 bytes --]

On Wed, Apr 06 2011, Ted Zlatanov wrote:

> I wonder if there's an easier way, maybe even with language completion
> or auto-guessing based on the original buffer's mode, that doesn't
> involve Org functions?  IIRC Julien posted how to do it with Org.

I'm not sure I understand what you want to do. Could you elaborate?

-- 
Julien Danjou
❱ http://julien.danjou.info

[-- Attachment #2: Type: application/pgp-signature, Size: 835 bytes --]

Re: quoting code blocks in Gnus without Org functions

[Ted Zlatanov]

On Wed, 06 Apr 2011 16:14:13 +0200 Julien Danjou <julien@danjou.info> wrote: 

JD> On Wed, Apr 06 2011, Ted Zlatanov wrote:
>> I wonder if there's an easier way, maybe even with language completion
>> or auto-guessing based on the original buffer's mode, that doesn't
>> involve Org functions?  IIRC Julien posted how to do it with Org.

JD> I'm not sure I understand what you want to do. Could you elaborate?

I want to copy some text into the clipboard and then paste it into an
Emacs buffer (not necessarily Gnus).  The original mode, e.g. lisp-mode,
should be considered when specifying the language of the code block.  So
if I copy my function, it should look like this when pasted:

#+begin_src lisp
(define-skeleton skeleton-src-quote
  "Insert a quoted Gnus source listing" "language: "
  \n
  > "#+begin_src " str \n
  > "(CODE GOES HERE)" \n
  > "#+end_src" \n \n)
#+end_src

...and I'd like to do it all without using Org functionality.  Is that
possible?

Ted

Re: quoting code blocks in Gnus without Org functions

[Julien Danjou]

[-- Attachment #1: Type: text/plain, Size: 1387 bytes --]

On Wed, Apr 06 2011, Ted Zlatanov wrote:

> I want to copy some text into the clipboard and then paste it into an
> Emacs buffer (not necessarily Gnus).  The original mode, e.g. lisp-mode,
> should be considered when specifying the language of the code block.  So
> if I copy my function, it should look like this when pasted:
>
> #+begin_src lisp
> (define-skeleton skeleton-src-quote
>   "Insert a quoted Gnus source listing" "language: "
>   \n
>   > "#+begin_src " str \n
>   > "(CODE GOES HERE)" \n
>   > "#+end_src" \n \n)
> #+end_src
>
> ...and I'd like to do it all without using Org functionality.  Is that
> possible?

Well, Org is only used to colorize the text rendered in Gnus, something
which is transparent since Org is included in Emacs.

What you want to do seems totally doable if we can have some information
about which buffer comes the yank from. But I doubt this is the case.

However you can still write something like (pseudo code :-)) and bind it
to a key:

(defun gnus-yank-code (code)
  (interactive (get-selection-content-if-active)
  (let ((buf (choose-a-buffer-using-something-like-gnus-dired-mail-buffers))
        (mode (symbol-name major-mode)))
    (with-current-buffer buf
      (insert "#+begin_src " mode "\n")
      (insert code)
      (insert "+end_src")))))

-- 
Julien Danjou
❱ http://julien.danjou.info

[-- Attachment #2: Type: application/pgp-signature, Size: 835 bytes --]

Re: quoting code blocks in Gnus without Org functions

[Ted Zlatanov]

On Wed, 06 Apr 2011 16:45:58 +0200 Julien Danjou <julien@danjou.info> wrote: 

JD> What you want to do seems totally doable if we can have some information
JD> about which buffer comes the yank from. But I doubt this is the case.

Here's my attempt.  This works for ELisp code but not for others (the
`cond' needs to be extended obviously).  Am I on the right track?  Is
there a map of mode to "short name" that the highlight code uses?  Or
can it just take the `major-mode' as string in addition to the "short
name", since it uses that for fontification anyhow?

(It felt very meta to copy the function with the function itself :)

#+begin_src lisp
(defun tzz-copy-region-with-mode-property (beg end)
  (interactive "r")
  (let ((text (buffer-substring beg end))
        (mode major-mode))
    (with-temp-buffer
      (insert text)
      (goto-char (point-min))
      (insert "#+begin_src "
              (cond
               ((memq mode '(lisp-interaction-mode emacs-lisp-mode))
                "lisp")
               (t "text"))
              "\n")
      (goto-char (point-max))
      (insert "\n"
              "#+end_src"
              "\n")
      (copy-region-as-kill (point-min) (point-max)))))
#+end_src

Thanks
Ted

Re: quoting code blocks in Gnus without Org functions

[Julien Danjou]

[-- Attachment #1: Type: text/plain, Size: 587 bytes --]

On Wed, Apr 06 2011, Ted Zlatanov wrote:

> Here's my attempt.  This works for ELisp code but not for others (the
> `cond' needs to be extended obviously).  Am I on the right track?  Is
> there a map of mode to "short name" that the highlight code uses?  Or
> can it just take the `major-mode' as string in addition to the "short
> name", since it uses that for fontification anyhow?

You do not need a cond. Simply put the current major mode symbol's name
without the -mode, e.g. lisp-interaction-mode -> lisp-interaction.

-- 
Julien Danjou
❱ http://julien.danjou.info

[-- Attachment #2: Type: application/pgp-signature, Size: 835 bytes --]

Re: quoting code blocks in Gnus without Org functions

[Ted Zlatanov]

On Thu, 07 Apr 2011 09:38:48 +0200 Julien Danjou <julien@danjou.info> wrote: 

JD> On Wed, Apr 06 2011, Ted Zlatanov wrote:
>> Here's my attempt.  This works for ELisp code but not for others (the
>> `cond' needs to be extended obviously).  Am I on the right track?  Is
>> there a map of mode to "short name" that the highlight code uses?  Or
>> can it just take the `major-mode' as string in addition to the "short
>> name", since it uses that for fontification anyhow?

JD> You do not need a cond. Simply put the current major mode symbol's name
JD> without the -mode, e.g. lisp-interaction-mode -> lisp-interaction.

Like this?  Or should I say "text" as a fallback?

#+begin_src emacs-lisp
(defun tzz-copy-region-with-mode-property (beg end)
  (interactive "r")
  (let ((text (buffer-substring beg end))
        (mode (symbol-name major-mode)))
    (with-temp-buffer
      (insert text)
      (goto-char (point-min))
      (insert "#+begin_src"
              (if (string-match "\\(.+\\)-mode" mode)
                  (concat " " (match-string 1 mode))
                "")
              "\n")
      (goto-char (point-max))
      (insert "\n"
              "#+end_src"
              "\n")
      (copy-region-as-kill (point-min) (point-max)))))

#+end_src

Thanks
Ted

Re: quoting code blocks in Gnus without Org functions

[Ted Zlatanov]

On Thu, 07 Apr 2011 14:57:41 -0500 Ted Zlatanov <tzz@lifelogs.com> wrote: 

TZ> On Thu, 07 Apr 2011 09:38:48 +0200 Julien Danjou <julien@danjou.info> wrote: 

JD> You do not need a cond. Simply put the current major mode symbol's name
JD> without the -mode, e.g. lisp-interaction-mode -> lisp-interaction.

TZ> #+begin_src emacs-lisp

This doesn't syntax-highlight like "lisp" does.  So "lisp" is still
better currently.

Ted

Re: quoting code blocks in Gnus without Org functions

[Ted Zlatanov]

On Thu, 07 Apr 2011 16:20:24 -0500 Ted Zlatanov <tzz@lifelogs.com> wrote: 

TZ> On Thu, 07 Apr 2011 14:57:41 -0500 Ted Zlatanov <tzz@lifelogs.com> wrote: 
TZ> On Thu, 07 Apr 2011 09:38:48 +0200 Julien Danjou <julien@danjou.info> wrote: 

JD> You do not need a cond. Simply put the current major mode symbol's name
JD> without the -mode, e.g. lisp-interaction-mode -> lisp-interaction.

TZ> #+begin_src emacs-lisp

TZ> This doesn't syntax-highlight like "lisp" does.  So "lisp" is still
TZ> better currently.

Huh, "lisp" is also not highlighted correctly in Gnus.  I distinctly
recall the Emacs Lisp highlighting was applied to some examples you
posted, Julien.  Is my example wrong or have I misconfigured something?

Thanks
Ted

Re: quoting code blocks in Gnus without Org functions

[Balázs Tóth]

Ted Zlatanov <tzz@lifelogs.com> writes:

> On Thu, 07 Apr 2011 16:20:24 -0500 Ted Zlatanov <tzz@lifelogs.com> wrote: 
>
> TZ> On Thu, 07 Apr 2011 14:57:41 -0500 Ted Zlatanov <tzz@lifelogs.com> wrote: 
> TZ> On Thu, 07 Apr 2011 09:38:48 +0200 Julien Danjou <julien@danjou.info> wrote: 
>
> JD> You do not need a cond. Simply put the current major mode symbol's name
> JD> without the -mode, e.g. lisp-interaction-mode -> lisp-interaction.
>
> TZ> #+begin_src emacs-lisp
>
> TZ> This doesn't syntax-highlight like "lisp" does.  So "lisp" is still
> TZ> better currently.
>
> Huh, "lisp" is also not highlighted correctly in Gnus.  I distinctly
> recall the Emacs Lisp highlighting was applied to some examples you
> posted, Julien.  Is my example wrong or have I misconfigured something?
>
> Thanks
> Ted

Both "lisp" and "emacs-lisp" version is highlighted nicely for me.

Regards,
Balazs

Re: quoting code blocks in Gnus without Org functions

[Julien Danjou]

[-- Attachment #1: Type: text/plain, Size: 1243 bytes --]

On Thu, Apr 07 2011, Ted Zlatanov wrote:

> Like this?  Or should I say "text" as a fallback?
>
> #+begin_src emacs-lisp
> (defun tzz-copy-region-with-mode-property (beg end)
>   (interactive "r")
>   (let ((text (buffer-substring beg end))
>         (mode (symbol-name major-mode)))
>     (with-temp-buffer
>       (insert text)
>       (goto-char (point-min))
>       (insert "#+begin_src"
>               (if (string-match "\\(.+\\)-mode" mode)
>                   (concat " " (match-string 1 mode))
>                 "")
>               "\n")
>       (goto-char (point-max))
>       (insert "\n"
>               "#+end_src"
>               "\n")
>       (copy-region-as-kill (point-min) (point-max)))))

You do not need a fallback AFAIK, you would just end without any
fontification. :)

> This doesn't syntax-highlight like "lisp" does.  So "lisp" is still
> better currently.

It does here.

> Huh, "lisp" is also not highlighted correctly in Gnus.  I distinctly
> recall the Emacs Lisp highlighting was applied to some examples you
> posted, Julien.  Is my example wrong or have I misconfigured something?

Check `org-src-fontify-natively' value?

-- 
Julien Danjou
❱ http://julien.danjou.info

[-- Attachment #2: Type: application/pgp-signature, Size: 835 bytes --]

Re: quoting code blocks in Gnus without Org functions

[Ted Zlatanov]

On Fri, 08 Apr 2011 11:40:17 +0200 Julien Danjou <julien@danjou.info> wrote: 

JD> Check `org-src-fontify-natively' value?

Strange, it's nil.  I could swear it was t before.  Thanks.

Ted

Re: draft group weirdness

[Eric Abrahamsen]

On Wed, Apr 06 2011, Eric Abrahamsen wrote:

> On Wed, Apr 06 2011, Lars Magne Ingebrigtsen wrote:
>
>> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>>
>>> I've been having a lot of trouble with the drafts group recently. I
>>> save drafts, go back into the drafts folder, hit 'e' to edit a message,
>>> and it asks me for a file to save into.
>>
>> That's bizarre.  (setq debug-on-quit t) and `C-g' when it prompts, so
>> that we can see what function you're calling with `e'.
>
> Here's the backtrace:
>
> #+begin_src lisp
> Debugger entered--Lisp error: (quit)
>   ad-Orig-completing-read("File to save in: ~/" (("dummy" . 1)) nil nil nil ido-file-history nil nil)
>   completing-read("File to save in: ~/" (("dummy" . 1)) nil nil nil ido-file-history)
>   byte-code("\304\305\b	\"\306\307\211\n\307\v&\207" [item prompt ido-text-init history completing-read ido-make-prompt (("dummy" . 1)) nil] 7)
>   ido-read-internal(file "File to save in: " ido-file-history nil nil nil)
>   ido-read-file-name("File to save in: " "~/" nil nil nil nil)
>   read-file-name("File to save in: ")
>   basic-save-buffer()
>   save-buffer()
>   gnus-draft-edit-message()
>   call-interactively(gnus-draft-edit-message nil nil)
> #+end_src

Other stuff…

My select method is nntp, and my secondary select methods nnml and nntp
(for gwene). Drafts get saved in ~/gnus/News/drafts/drafts, though there
is also an empty ~/gnus/Mail/drafts directory as well. Editing a single
saved earlier draft creates not only a "1" file in the above drafts
directory, but also a "2" and "3", all with the same content. These
files are not erased after the original message is sent or deleted.
(They might get erased later, I'm not sure.)

Losing marks happens when I am replying to an existing message, and save
that reply as a draft. Then when I edit the draft (or perhaps when I
send it?), the group that the original is in loses all its marks.

Hopefully something in there will raise a red flag. I've gotten better
at the debugger; "save-buffer" is obviously getting called at line 103
of gnus-draft.el, maybe I could figure out why that's the wrong buffer…

Thanks,
Eric

Re: draft group weirdness

[Ted Zlatanov]

On Sun, 10 Apr 2011 16:50:53 +0100 Eric Abrahamsen <eric@ericabrahamsen.net> wrote: 

>> Here's the backtrace:
>> 
>> #+begin_src lisp
>> Debugger entered--Lisp error: (quit)

Hmm, either the article highlighting should be changed to detect quoted
source, or quoted source should be excluded from quoting rules.

Ted

Re: draft group weirdness

[Eric Abrahamsen]

On Wed, Apr 06 2011, Eric Abrahamsen wrote:

> On Wed, Apr 06 2011, Lars Magne Ingebrigtsen wrote:
>
>> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>>
>>> I've been having a lot of trouble with the drafts group recently. I
>>> save drafts, go back into the drafts folder, hit 'e' to edit a message,
>>> and it asks me for a file to save into.
>>

Also! Just realized that the extra copies of the drafts do not get
automatically deleted, they popular the drafts folder and I have to
delete them by hand. My guess is the culprit is in here somewhere:

(setq gnus-home-directory "~/gnus/"
      gnus-directory "~/gnus/News/"
      message-directory "~/gnus/Mail/"
      nnmail-message-id-cache-file "~/gnus/.nnmail-cache"
      message-auto-save-directory "~/gnus/Mail/drafts"
      gnus-cache-directory "~/gnus/News/cache")

Note that the "message-auto-save-directory" is *not* where messages get
autosaved, they go into the ~/gnus/News/drafts/drafts directory, same
with the deliberately saved messages.

Eric

Re: draft group weirdness

[Lars Magne Ingebrigtsen]

Eric Abrahamsen <eric@ericabrahamsen.net> writes:

> Also! Just realized that the extra copies of the drafts do not get
> automatically deleted, they popular the drafts folder and I have to
> delete them by hand. My guess is the culprit is in here somewhere:
>
> (setq gnus-home-directory "~/gnus/"
>       gnus-directory "~/gnus/News/"
>       message-directory "~/gnus/Mail/"
>       nnmail-message-id-cache-file "~/gnus/.nnmail-cache"
>       message-auto-save-directory "~/gnus/Mail/drafts"
>       gnus-cache-directory "~/gnus/News/cache")

Ah, right.  Well, something is confusing Gnus about...  something.  :-)

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

Re: draft group weirdness

[Eric Abrahamsen]

On Tue, Apr 12 2011, Lars Magne Ingebrigtsen wrote:

> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>
>> Also! Just realized that the extra copies of the drafts do not get
>> automatically deleted, they popular the drafts folder and I have to
>> delete them by hand. My guess is the culprit is in here somewhere:
>>
>> (setq gnus-home-directory "~/gnus/"
>>       gnus-directory "~/gnus/News/"
>>       message-directory "~/gnus/Mail/"
>>       nnmail-message-id-cache-file "~/gnus/.nnmail-cache"
>>       message-auto-save-directory "~/gnus/Mail/drafts"
>>       gnus-cache-directory "~/gnus/News/cache")
>
> Ah, right.  Well, something is confusing Gnus about...  something.  :-)

I came to the same conclusion! :)

I've done a fair amount of setting things up one way and then changing
them to another, and I think some of the changes have not been "clean".
I'd be perfectly happy to blast my setup and start afresh, if I was
confident that I'd be able to really clean out all the vestiges of prior
settings…

Re: draft group weirdness

[Lars Magne Ingebrigtsen]

Eric Abrahamsen <eric@ericabrahamsen.net> writes:

>   read-file-name("File to save in: ")
>   basic-save-buffer()
>   save-buffer()
>   gnus-draft-edit-message()

I've been trying to debug this, but without any major success.  The
symptom that's triggering this is that `buffer-file-name' is nil when
you hit `e'.  That doesn't happen for me.  Do you know whether you have
any customisations that would make your buffer file names disappear?  (I
know, that's like looking for a needle in a haystack...)

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

Re: draft group weirdness

[Eric Abrahamsen]

Eric Abrahamsen <eric@ericabrahamsen.net> writes:

> On Wed, Apr 06 2011, Lars Magne Ingebrigtsen wrote:
>
>> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>>
>>> I've been having a lot of trouble with the drafts group recently. I
>>> save drafts, go back into the drafts folder, hit 'e' to edit a message,
>>> and it asks me for a file to save into.
>>
>> That's bizarre.  (setq debug-on-quit t) and `C-g' when it prompts, so
>> that we can see what function you're calling with `e'.

I think I found the culprit here: window configuration. Doesn't make a
whole lot of sense, but neither did the error to begin with. I had a
three-pane thing going on, and I recently got rid of it because it was
such a pain to manually move point to the composition buffer when
replying to messages.

After I went back to the default, my draft strangeness went away.
Editing a draft multiple times still creates multiple editing buffers
for that message, but maybe that's normal.

It also cleared up another, equally weird problem: my mairix search
results folder was appearing in random places in the filesystem. My org
directory, recent directories I'd been working in, other places gnus had
no business knowing about… My eventual sense was that my last 'pwd'
location before switching to gnus was somehow hanging around and getting
substituted for gnus' base directory. This is mostly useless as a bug
report, I know, but I can't help mentioning it.

It's all very magical, but that's why I love emacs/gnus…

Eric