Is this a buglet.. nndrafts:draft renegades

Is this a buglet.. nndrafts:draft renegades

[Harry Putnam]

I keep finding messages in nndraft:drafts that I didn't put there.
They get there by auto saving and gnus code I guess.  But if the
original message is killed (C-x k) before mailing and with having
never pressed C-x C-f or C-x C-d then that draft should die too right?

It doesn't under some circumstances.  Detailed here inside of just
such a message.  I started a composition then left it for a time with
C-x b.  Came back later, continued finally deciding to C-x k the whole
thing.

Next fetch I see a `1' light up nndrafts:draft . Take a look and
there is a partial copy of the killed messages.  It's detailed in the
file below (Some headers snipped to shorten it up)

C-x i t/killed_mail_message
======================================== 
  X-Draft-From: ("comp.editors" 38793)
  Newsgroups: comp.editors
  Subject: Re: Elvis 2.2f-beta is now available
  From: Harry Putnam <reader@newsguy.com>
  Date: Fri, 02 Nov 2001 11:35:12 -0800
  Message-ID: <m1zo65ezkf.fsf@reader.newsguy.com>
  --text follows this line--
  
  Starting this followup:
  Fri Nov  2 11:35:21 PST 2001
  
  I think I keep finding messages in nndraft:drafts that should not be
  there.  It works like this:
  Start a message of which this is an example.  Work on it for a while,
  then leave it temporarily for a while with C-x b *Group* or similar.
  
  Fri Nov  2 11:37:20 PST 2001
  Doing that very thing now
  
  ========================================
  
  Fri Nov 2 12:53:32 PST 2001 
  
  Here we is .. back about an hour and 20 min later.  Now I'm going to
  append-to-file this buffer then C-h k kill it.  Look in nndraft:drafts
  and see if something is there.  My feeling is there should not be.
  
  Waiting for auto save
  
  Fri Nov  2 12:55:49 PST 2001
  Auto save just occured appending then killing
  
  ......

Fri Nov  2 12:58:52 PST 2001

Look what I found in nndraft:drafts.
This shouldn't be here right?

========================================

  X-Draft-From: ("comp.editors" 38793)
  Newsgroups: comp.editors
  Subject: Re: Elvis 2.2f-beta is now available
  From: Harry Putnam <reader@newsguy.com>
  Date: Fri, 02 Nov 2001 11:35:12 -0800
  Message-ID: <m1zo65ezkf.fsf@reader.newsguy.com>
  
  
  Starting this followup:
  Fri Nov  2 11:35:21 PST 2001
  
  I think I keep finding messages in nndraft:drafts that should not be
  there.  It works like this:
  Start a message of which this is an example.  Work on it for a while,
  then leave it temporarily for a while with C-x b *Group* or similar.
  
  Fri Nov  2 11:37:20 PST 2001
  Doing that very thing now

Re: Is this a buglet.. nndrafts:draft renegades

[Matt Armstrong]

Harry Putnam <reader@newsguy.com> writes:

> I keep finding messages in nndraft:drafts that I didn't put there.
> They get there by auto saving and gnus code I guess.  But if the
> original message is killed (C-x k) before mailing and with having
> never pressed C-x C-f or C-x C-d then that draft should die too right?

Let's see, I'm going to C-x s and then C-x k this buffer.  Then I'll
check if it is in the drafts group.

Yes it was.

This seems to be a feature, not a bug.  ;-) Once saved to disk, a
draft is probably best deleted with B DEL from the summary buffer.

Re: Is this a buglet.. nndrafts:draft renegades

[Harry Putnam]

"Matt Armstrong" <matt+dated+1007329560.f6b614@lickey.com> writes:

> Harry Putnam <reader@newsguy.com> writes:
>
>> I keep finding messages in nndraft:drafts that I didn't put there.
>> They get there by auto saving and gnus code I guess.  But if the
>> original message is killed (C-x k) before mailing and with having
>> never pressed C-x C-f or C-x C-d then that draft should die too right?

C-x C-f above was a typo  .. should have said C-x C-s

> Let's see, I'm going to C-x s and then C-x k this buffer.  Then I'll
> check if it is in the drafts group.

> Yes it was.

You may notice the very first line `I didn' put there'
That is, no C-x s was run unless that is run transparently at some
point.

The behavior I describe above is new.  Since maybe a few week.

> This seems to be a feature, not a bug.  ;-) Once saved to disk, a
> draft is probably best deleted with B DEL from the summary buffer.

It was never saved to disk.  Only auto saving occured as detailed in
the message.  Not sure where you are getting the idea something has
been saved.  Maybe you misread C-x b, or maybe my typo threw you off.

I'm saying this message was never overtly saved in any way.

Only insofar as gnus keeps a running copy in drafts and auto saving
occuring. 

So start a composition... leave it with C-x b (switch buffers)
Come back later and maybe continue the composition.  Finally with no
overt save taking place C-x k (kill the buffer).

That message should not show up in drafts..... at least it didn't used to.

Re: Is this a buglet.. nndrafts:draft renegades

[Russ Allbery]

Matt Armstrong <matt+dated+1007329560.f6b614@lickey.com> writes:
> Harry Putnam <reader@newsguy.com> writes:

>> I keep finding messages in nndraft:drafts that I didn't put there.
>> They get there by auto saving and gnus code I guess.  But if the
>> original message is killed (C-x k) before mailing and with having never
>> pressed C-x C-f or C-x C-d then that draft should die too right?

> Let's see, I'm going to C-x s and then C-x k this buffer.  Then I'll
> check if it is in the drafts group.

> Yes it was.

> This seems to be a feature, not a bug.  ;-)

Yeah, that's how I always put stuff into drafts.  What's the right way?
The manual for 5.8.7 strongly implies that's the correct way too.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Is this a buglet.. nndrafts:draft renegades

[Harry Putnam]

Russ Allbery <rra@stanford.edu> writes:

> Matt Armstrong <matt+dated+1007329560.f6b614@lickey.com> writes:
>> Harry Putnam <reader@newsguy.com> writes:
>
>>> I keep finding messages in nndraft:drafts that I didn't put there.
>>> They get there by auto saving and gnus code I guess.  But if the
>>> original message is killed (C-x k) before mailing and with having never
>>> pressed C-x C-f or C-x C-d then that draft should die too right?
>
>> Let's see, I'm going to C-x s and then C-x k this buffer.  Then I'll
>> check if it is in the drafts group.
>
>> Yes it was.
>
>> This seems to be a feature, not a bug.  ;-)
>
> Yeah, that's how I always put stuff into drafts.  What's the right way?
> The manual for 5.8.7 strongly implies that's the correct way too.

The point here, at least my point, was that when no way is used then
nothing should stay in drafts.

Re: Is this a buglet.. nndrafts:draft renegades

[Karl Kleinpaste]

Russ Allbery <rra@stanford.edu> writes:
> Yeah, that's how I always put stuff into drafts.  What's the right way?

`C-c C-d' runs message-dont-send, which files away the message into
drafts, then destroys the buffer.

As opposed to `C-c C-k', which simply destroys the buffer.

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

Karl Kleinpaste <karl@charcoal.com> writes:

> `C-c C-d' runs message-dont-send, which files away the message into
> drafts, then destroys the buffer.
>
> As opposed to `C-c C-k', which simply destroys the buffer.

When I run C-c C-k, the message is still filed into my drafts.  Is
this a bug then?

I've brought this up twice before on ding, and no one's responded :-(.

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Russ Allbery]

Karl Kleinpaste <karl@charcoal.com> writes:
> Russ Allbery <rra@stanford.edu> writes:

>> Yeah, that's how I always put stuff into drafts.  What's the right way?

> `C-c C-d' runs message-dont-send, which files away the message into
> drafts, then destroys the buffer.

Could someone include that information in the docstring for C-c C-d and in
the manual section under Drafts?  The docstring at least in 5.8.7 says
nothing at all about drafts, and therefore sort of implies that the
message just goes away entirely.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Is this a buglet.. nndrafts:draft renegades

[Per Abrahamsen]

Harry Putnam <reader@newsguy.com> writes:

> That message should not show up in drafts..... at least it didn't used to.

I think the old behaviour was because of a bug.  

"Drafts" is where your message is until you either send it (e.g. with
C-c C-c) or kill it (e.g. with C-c C-k).

That can safe a lot of work if Emacs dies while you compose a long
message.

Re: Is this a buglet.. nndrafts:draft renegades

[Per Abrahamsen]

Samuel Padgett <spadgett1@nc.rr.com> writes:

> When I run C-c C-k, the message is still filed into my drafts.  Is
> this a bug then?

Yes, I believe so.

> I've brought this up twice before on ding, and no one's responded :-(.

It works for me.  What is C-c C-k bound to?

Re: Is this a buglet.. nndrafts:draft renegades

[ShengHuo ZHU]

Per Abrahamsen <abraham@dina.kvl.dk> writes:

> Samuel Padgett <spadgett1@nc.rr.com> writes:
>
>> When I run C-c C-k, the message is still filed into my drafts.  Is
>> this a bug then?
>
> Yes, I believe so.

I knew the bug.  If most users think that it is bad, I can fix it
later today or tomorrow.

>> I've brought this up twice before on ding, and no one's responded :-(.
>
> It works for me.  What is C-c C-k bound to?

The binding is fine. Just need to add a kill-action.  

ShengHuo

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

Per Abrahamsen <abraham@dina.kvl.dk> writes:

> Samuel Padgett <spadgett1@nc.rr.com> writes:
>
>> When I run C-c C-k, the message is still filed into my drafts.?

[...]

> It works for me.  What is C-c C-k bound to?

"C-c C-k runs the command message-kill-buffer"

Thanks,
Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

ShengHuo ZHU <zsh@cs.rochester.edu> writes:

> Per Abrahamsen <abraham@dina.kvl.dk> writes:
>
>> Samuel Padgett <spadgett1@nc.rr.com> writes:
>>
>>> When I run C-c C-k, the message is still filed into my drafts.  Is
>>> this a bug then?
>>
>> Yes, I believe so.
>
> I knew the bug.  If most users think that it is bad, I can fix it
> later today or tomorrow.

This started happening when Lars installed some magic to make autosave
files go to the drafts folder on an Emacs crash.  I very much like
this behavior and don't want it to change.  When I run an explicit
command to kill the message (like C-c C-k), however, it's a bit of a
nuisance to have to go delete the message in the drafts group as well.
I don't think the message so go there on an explicit kill with C-c
C-k.

Thanks, ShengHuo.

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Kai Großjohann]

Russ Allbery <rra@stanford.edu> writes:

> Yeah, that's how I always put stuff into drafts.  What's the right way?
> The manual for 5.8.7 strongly implies that's the correct way too.

I use C-c C-d to put things in nndraft:drafts.

kai
-- 
Lisp is kinda like tpircstsoP

Re: Is this a buglet.. nndrafts:draft renegades

[Kai Großjohann]

Russ Allbery <rra@stanford.edu> writes:

> Could someone include that information in the docstring for C-c C-d and in
> the manual section under Drafts?  The docstring at least in 5.8.7 says
> nothing at all about drafts, and therefore sort of implies that the
> message just goes away entirely.

Hm.  Indeed.  There is a doc bug here.  It should mention that it
also runs the actions listed in `message-postpone-actions'.  Users
who look there can see the relevant actions for Gnus.

Hm.  But for me, that's only frobbing the window config.  Hm.

kai
-- 
Lisp is kinda like tpircstsoP

Re: Is this a buglet.. nndrafts:draft renegades

[Matt Armstrong]

Harry Putnam <reader@newsguy.com> writes:

>> This seems to be a feature, not a bug.  ;-) Once saved to disk, a
>> draft is probably best deleted with B DEL from the summary buffer.
>
> So start a composition... leave it with C-x b (switch buffers) Come
> back later and maybe continue the composition.  Finally with no
> overt save taking place C-x k (kill the buffer).

I agree that if you C-x k the buffer, it shouldn't end up in the
drafts buffer.


ShengHuo ZHU <zsh@cs.rochester.edu> writes:

> I knew the bug.  If most users think that it is bad, I can fix it
> later today or tomorrow.
>
>>> I've brought this up twice before on ding, and no one's responded :-(.
>>
>> It works for me.  What is C-c C-k bound to?
>
> The binding is fine. Just need to add a kill-action.  

Fine by me, and Harry too.  ;-)


-- 
matt

Re: Is this a buglet.. nndrafts:draft renegades

[Russ Allbery]

Per Abrahamsen <abraham@dina.kvl.dk> writes:

> I think the old behaviour was because of a bug.  

> "Drafts" is where your message is until you either send it (e.g. with
> C-c C-c) or kill it (e.g. with C-c C-k).

> That can safe a lot of work if Emacs dies while you compose a long
> message.

I'm also having trouble with this part of the system.  The autosave files
don't seem to ever show up in Drafts, using Gnus 5.8.7 and XEmacs
21.1.12.  I have to explicitly save at least once with C-x C-s, or there's
no trace of the message when XEmacs crashes.

(Once I explicitly save once, that saved draft does seem to get any
autosave updates; in other words, I don't have to continously save again
as the message changes.  This makes me think that it's not a problem with
autosave but instead somehow a problem with the way that Gnus assigns a
save file to autosave.)

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

Re: Is this a buglet.. nndrafts:draft renegades

[Matt Armstrong]

"Matt Armstrong" <matt+dated+1007407971.e3d11c@lickey.com> writes:

> I agree that if you C-x k the buffer, it shouldn't end up in the
> drafts buffer.

Actually, I reverse my opinion.  C-x k deletes a buffer, not a file.

I have C-x C-s C-x k deeply ingrained in my muscle memory -- it is the
"save to disk, then delete buffer" key sequence.  Data does not get
lost when doing this.

If C-x k suddenly deletes the message out of the drafts folder, I'll
end up accidentally deleting drafts left and right.


> ShengHuo ZHU <zsh@cs.rochester.edu> writes:
>
>> I knew the bug.  If most users think that it is bad, I can fix it
>> later today or tomorrow.
>>
>>>> I've brought this up twice before on ding, and no one's responded
>>>> :-(.
>>>
>>> It works for me.  What is C-c C-k bound to?
>>
>> The binding is fine. Just need to add a kill-action.
>
> Fine by me, and Harry too.  ;-)

Not fine by me any more.

I like the idea that a message I'm editing is actually a file, and
that if it saved (via auto-save or not), it is around until I
explicitly delete it.

I'd be fine if message buffers prompted me "delete message from drafts
folder too?" when they were killed.  Or we could provide a "kill this
message buffer and its draft" key binding.


-- 
matt

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

"Matt Armstrong" <matt+dated+1007413841.964140@lickey.com> writes:

> I have C-x C-s C-x k deeply ingrained in my muscle memory -- it is the
> "save to disk, then delete buffer" key sequence.  Data does not get
> lost when doing this.
>
> If C-x k suddenly deletes the message out of the drafts folder, I'll
> end up accidentally deleting drafts left and right.

But we aren't talking about C-x k.  We're talking about C-c C-k,
right?

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Harry Putnam]

"Matt Armstrong" <matt+dated+1007413841.964140@lickey.com> writes:

[...]

> Actually, I reverse my opinion.  C-x k deletes a buffer, not a file.
>
> I have C-x C-s C-x k deeply ingrained in my muscle memory -- it is the
> "save to disk, then delete buffer" key sequence.  Data does not get
> lost when doing this.

In the sense that a composition is a file, C-x k does kill the file
and data is lost in every other usage.  However it does prompt for the
permission. 

The sequence below assumes `file' does not yet exist.  Just like a
fresh composition buffer.
C-x C-f file 
type Start
C-x k  prompt: yes 
The file is removed and data is lost.   Why should a composition
buffer be any different?

>
> If C-x k suddenly deletes the message out of the drafts folder, I'll
> end up accidentally deleting drafts left and right.

Why is this so perilous compared to behavior on a not existent file in
C-x C-f file?

> I'd be fine if message buffers prompted me "delete message from drafts
> folder too?" when they were killed.  Or we could provide a "kill this
> message buffer and its draft" key binding.

Currently C-x k does prompt for permission to continue.  Maybe you
have it turned off.

In this buffer I press C-x k and get:
Kill buffer: (default *wide reply to Matt Armstrong*) 

Do you need more than that? 

Re: Is this a buglet.. nndrafts:draft renegades

[Harry Putnam]

Samuel Padgett <spadgett1@nc.rr.com> writes:

> "Matt Armstrong" <matt+dated+1007413841.964140@lickey.com> writes:
>
>> I have C-x C-s C-x k deeply ingrained in my muscle memory -- it is the
>> "save to disk, then delete buffer" key sequence.  Data does not get
>> lost when doing this.
>>
>> If C-x k suddenly deletes the message out of the drafts folder, I'll
>> end up accidentally deleting drafts left and right.
>
> But we aren't talking about C-x k.  We're talking about C-c C-k,
> right?

I was talking about C-x k
I'd kind of forgotten about C-c C-k.  That does kill the Drafts/copy
here.  I guess that is what I should be using.  I still think C-x k
should do it to... to be consisitent. 

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

Harry Putnam <reader@newsguy.com> writes:

>> But we aren't talking about C-x k.  We're talking about C-c C-k,
>> right?
>
> I was talking about C-x k

Oh, sorry.

> I'd kind of forgotten about C-c C-k.  That does kill the Drafts/copy
> here.

Hm.  If I compose a message from the *Group* buffer using "m" and then
type "C-c C-k" midway through, it is put in my drafts folder.  Is this
not what you see?

> I guess that is what I should be using.  I still think C-x k should
> do it to... to be consisitent.

This would be _inconsistent_ with the way C-x k works in the rest of
Emacs.  I only think the message should not go to the drafts folder
after C-x k if it's unmodified and hasn't been saved.  After all,
isn't this what C-c C-k is for: to cancel the message without making
it a draft?  And if this is not what C-c C-k is for, why have both C-c
C-k and C-c C-d?

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Harry Putnam]

Samuel Padgett <spadgett1@nc.rr.com> writes:

[...]

> Hm.  If I compose a message from the *Group* buffer using "m" and then
> type "C-c C-k" midway through, it is put in my drafts folder.  Is this
> not what you see?
>
>> I guess that is what I should be using.  I still think C-x k should
>> do it to... to be consisitent.

> This would be _inconsistent_ with the way C-x k works in the rest of
> Emacs.  I only think the message should not go to the drafts folder

How so?  m in gnus opens a fresh composition buffer where none was
before.  C-x k here and it lingers in drafts.

C-x C-f can open  a file where none existed before and present you with
a composition buffer. If you use C-x k here it all dies.  It doesn't
linger somewhere.

I was saying C-x k should be the same in gnus as everywhere else. In
fact it was.

> after C-x k if it's unmodified and hasn't been saved.  After all,
> isn't this what C-c C-k is for: to cancel the message without making
> it a draft?  And if this is not what C-c C-k is for, why have both C-c
> C-k and C-c C-d?

Yeah, I think I just need to switch to using C-c C-k.  Just my habit
is C-x k which has always worked until recently.

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

Harry Putnam <reader@newsguy.com> writes:

> Samuel Padgett <spadgett1@nc.rr.com> writes:
>
>> This would be _inconsistent_ with the way C-x k works in the rest of
>> Emacs.
>
> How so?  m in gnus opens a fresh composition buffer where none was
> before.  C-x k here and it lingers in drafts.

Yes.  This is the same thing that would happen if Emacs had crashed
when you were composing the message.  So Gnus behaves consistently
when there's an autosave file, regardless of whether the autosave is
there because of a crash or a C-x k.

> C-x C-f can open  a file where none existed before and present you with
> a composition buffer. If you use C-x k here it all dies.  It doesn't
> linger somewhere.

It does linger.  An autosave file will stay around if you kill a
modified buffer with C-x k, right?

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

Samuel Padgett <spadgett1@nc.rr.com> writes:

> So Gnus behaves consistently when there's an autosave file,
> regardless of whether the autosave is there because of a crash or a
> C-x k.

After thinking about this some more, I'm starting to change my mind.
C-c C-k should generate an autosave file as well.  Hm.

Maybe it's right that the message goes to the drafts folder.

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Kai Großjohann]

"Matt Armstrong" <matt+dated+1007413841.964140@lickey.com> writes:

> "Matt Armstrong" <matt+dated+1007407971.e3d11c@lickey.com> writes:
>
>> I agree that if you C-x k the buffer, it shouldn't end up in the
>> drafts buffer.
>
> Actually, I reverse my opinion.  C-x k deletes a buffer, not a file.
>
> I have C-x C-s C-x k deeply ingrained in my muscle memory -- it is the
> "save to disk, then delete buffer" key sequence.  Data does not get
> lost when doing this.

I suggest that C-x k delete the autosave file but not the real file.

The drafts group has two uses: first, autosave files are saved there
so that you can recover your half-finished messages after an Emacs
crash.  Second, C-x C-s and C-c C-d save stuff there.

What do you think?

kai
-- 
I like BOTH kinds of music.

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> "Matt Armstrong" <matt+dated+1007413841.964140@lickey.com> writes:
>
>> I have C-x C-s C-x k deeply ingrained in my muscle memory -- it is the
>> "save to disk, then delete buffer" key sequence.  Data does not get
>> lost when doing this.
>
> I suggest that C-x k delete the autosave file but not the real file.

But what if I accidentally kill a buffer while composing a message I
haven't saved?  There would be no way to recover the message.

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Kai Großjohann]

Samuel Padgett <spadgett1@nc.rr.com> writes:

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>
>> I suggest that C-x k delete the autosave file but not the real file.
>
> But what if I accidentally kill a buffer while composing a message I
> haven't saved?  There would be no way to recover the message.

Hm.

kai
-- 
I like BOTH kinds of music.

Re: Is this a buglet.. nndrafts:draft renegades

[Matt Armstrong]

Samuel Padgett <spadgett1@nc.rr.com> writes:

> Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:
>
>> "Matt Armstrong" <matt+dated+1007413841.964140@lickey.com> writes:
>>
>>> I have C-x C-s C-x k deeply ingrained in my muscle memory -- it is the
>>> "save to disk, then delete buffer" key sequence.  Data does not get
>>> lost when doing this.
>>
>> I suggest that C-x k delete the autosave file but not the real file.
>
> But what if I accidentally kill a buffer while composing a message I
> haven't saved?  There would be no way to recover the message.

Emacs will prompt you with "Buffer X modified; kill anyway?" -- I
think that is good enough.


-- 
matt

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

"Matt Armstrong" <matt+dated+1007493620.315b12@lickey.com> writes:

> Samuel Padgett <spadgett1@nc.rr.com> writes:
>
>> But what if I accidentally kill a buffer while composing a message I
>> haven't saved?  There would be no way to recover the message.
>
> Emacs will prompt you with "Buffer X modified; kill anyway?" -- I
> think that is good enough.

No, I disagree.  If I do this to a C source file, I can always recover
my changes using the autosave file.  But I can't now with a message I
was composing?  Why?

I usually don't worry too much when I'm asked the "Buffer modified;
kill anyway?" question because I know Emacs will always kindly leave
an autosave file for me just in case I'm doing something boneheaded.

C-x k is a global command.  Changing its meaning in message-mode--and
making it more destructive to boot--is a bad idea.

Maybe changing C-c C-k to delete the autosave file is OK, though.  I
wonder if anyone relies on the current behavior?

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Per Abrahamsen]

Harry Putnam <reader@newsguy.com> writes:

> In the sense that a composition is a file, C-x k does kill the file
> and data is lost in every other usage. 

Does `C-x k' really any already generated autosave file?  If so, I
agree it should do the same for composition buffers.  If not, I
believe it shouldn't do it for composition buffers either.

Re: Is this a buglet.. nndrafts:draft renegades

[Harry Putnam]

[ NOTE sorry if this proves to be redundant but somehow I am getting
an unusual behavior here.  Posted a message just like this one a few
minutes ago but it appeared in an unexpected format on ding.
See Message-ID: <m18zdlb5qh.fsf@totally-fudged-out-message-id>

(I think Samuel or someone already posted on this and its cvs cured
problem I hope.... I'll be cvs updating shortly)]

Per Abrahamsen <abraham@dina.kvl.dk> writes:

> Harry Putnam <reader@newsguy.com> writes:
>
>> In the sense that a composition is a file, C-x k does kill the file
>> and data is lost in every other usage. 
>
> Does `C-x k' really any already generated autosave file?  If so, I
> agree it should do the same for composition buffers.  If not, I
> believe it shouldn't do it for composition buffers either.

M-x cd 
test_dir <RET>

C-x C-f tnew

Type:
Testing the action of C-x k on new composition buffer of tnew
before an auto-save has occured
<C-x k yes>

C-x d <RET> (gone without a trace)

Type:
Testing the action of C-x k on new composition buffer of tnew
after an auto-save has occured
<waiting>
excerpt from *Nessages*
        Auto-saving...done

<C-x k yes>

C-x d <RET> g 
(cong without a trace)

That is the behavior I see.  And I agree that a new message
composition should behave the same.

Re: Is this a buglet.. nndrafts:draft renegades

[Matt Armstrong]

Harry Putnam <reader@newsguy.com> writes:

> "Matt Armstrong" <matt+dated+1007413841.964140@lickey.com> writes:
>
> [...]
>
>> Actually, I reverse my opinion.  C-x k deletes a buffer, not a file.
>>
>> I have C-x C-s C-x k deeply ingrained in my muscle memory -- it is the
>> "save to disk, then delete buffer" key sequence.  Data does not get
>> lost when doing this.
>
> In the sense that a composition is a file, C-x k does kill the file
> and data is lost in every other usage.  However it does prompt for
> the permission.

I think we agree.  C-x k does not delete the file from disk if it has
been explicitly saved -- it does get rid of any autosave file and that
is fine with me.

It works this way in the drafts folder too (perhaps only with recent
CVS though).


> The sequence below assumes `file' does not yet exist.  Just like a
> fresh composition buffer.
> C-x C-f file 
> type Start
> C-x k  prompt: yes 
> The file is removed and data is lost.   Why should a composition
> buffer be any different?

I agree that it shouldn't be different.


>> If C-x k suddenly deletes the message out of the drafts folder, I'll
>> end up accidentally deleting drafts left and right.
>
> Why is this so perilous compared to behavior on a not existent file
> in C-x C-f file?

I'm saying that after I save the message (C-x C-s), C-x k should not
delete it from drafts.  But I don't think you want that either.


-- 
matt

Re: Is this a buglet.. nndrafts:draft renegades

[Paul Jarc]

Samuel Padgett <spadgett1@nc.rr.com> wrote:
> "Matt Armstrong" <matt+dated+1007493620.315b12@lickey.com> writes:
>> Samuel Padgett <spadgett1@nc.rr.com> writes:
>>> But what if I accidentally kill a buffer while composing a message I
>>> haven't saved?  There would be no way to recover the message.
>>
>> Emacs will prompt you with "Buffer X modified; kill anyway?" -- I
>> think that is good enough.
>
> No, I disagree.  If I do this to a C source file, I can always recover
> my changes using the autosave file.

No, you can't.  That autosave file is deleted when you kill the
buffer.  At least, that's what I see.  Maybe this is a result of
something weird in my .emacs, but I'm not seeing it.

It will always be possible for you to shoot yourself in the foot.  Any
tool that makes that impossible also makes it impossible to do
anything useful.


paul

Re: Is this a buglet.. nndrafts:draft renegades

[Matt Armstrong]

prj@po.cwru.edu (Paul Jarc) writes:

> Samuel Padgett <spadgett1@nc.rr.com> wrote:
>> "Matt Armstrong" <matt+dated+1007493620.315b12@lickey.com> writes:
>>> Samuel Padgett <spadgett1@nc.rr.com> writes:
>>>> But what if I accidentally kill a buffer while composing a message I
>>>> haven't saved?  There would be no way to recover the message.
>>>
>>> Emacs will prompt you with "Buffer X modified; kill anyway?" -- I
>>> think that is good enough.
>>
>> No, I disagree.  If I do this to a C source file, I can always recover
>> my changes using the autosave file.
>
> No, you can't.  That autosave file is deleted when you kill the
> buffer.  At least, that's what I see.  Maybe this is a result of
> something weird in my .emacs, but I'm not seeing it.
>
> It will always be possible for you to shoot yourself in the foot.  Any
> tool that makes that impossible also makes it impossible to do
> anything useful.

Samuel, could you be thinking of the backup file and not the autosave
file?

The autosave file:

    - for file "foo" is called "#foo#"
    - contain recent edits that have not yet been saved
    - is deleted whenever you save or kill a buffer
    - if you open the "foo" file when "#foo#" exists on disk, you get
      warned about possibly wanting to run M-x recover-file.

The backup files:

    - for file "foo" are called "foo~"
    - contain the contents of the file before your most recent save
    - remain on disk whenever you save or kill a buffer

-- 
matt

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

prj@po.cwru.edu (Paul Jarc) writes:

> Samuel Padgett <spadgett1@nc.rr.com> wrote:
>
>> No, I disagree.  If I do this to a C source file, I can always recover
>> my changes using the autosave file.
>
> No, you can't.  That autosave file is deleted when you kill the
> buffer.  At least, that's what I see.  Maybe this is a result of
> something weird in my .emacs, but I'm not seeing it.

C-h v delete-auto-save-files RET:

    delete-auto-save-files's value is t

    Documentation:
    *Non-nil means delete auto-save file when a buffer is saved or killed.

    Note that auto-save file will not be deleted if the buffer is killed
    when it has unsaved changes.

    You can customize this variable.

    Defined in `files'.

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

"Matt Armstrong" <matt+dated+1007571874.3d8167@lickey.com> writes:

> I think we agree.  C-x k does not delete the file from disk if it has
> been explicitly saved -- it does get rid of any autosave file and that
> is fine with me.

C-x k should not delete the autosave file of a modified
buffer--unless the docstring for `delete-auto-save-files' is wrong.
Or unless I'm reading it wrong.  Does it for you?

Maybe this is an Emacs bug?

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

Harry Putnam <reader@newsguy.com> writes:

> Type:
> Testing the action of C-x k on new composition buffer of tnew
> after an auto-save has occured
> <waiting>
> excerpt from *Nessages*
>         Auto-saving...done
>
> <C-x k yes>
>
> C-x d <RET> g 
> (cong without a trace)

Was the buffer modified when you killed it?

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Paul Jarc]

Samuel Padgett <spadgett1@nc.rr.com> wrote:
> prj@po.cwru.edu (Paul Jarc) writes:
>> Samuel Padgett <spadgett1@nc.rr.com> wrote:
>>> No, I disagree.  If I do this to a C source file, I can always recover
>>> my changes using the autosave file.
>>
>> No, you can't.  That autosave file is deleted when you kill the
>> buffer.  At least, that's what I see.  Maybe this is a result of
>> something weird in my .emacs, but I'm not seeing it.
>
> C-h v delete-auto-save-files RET:
>
>     delete-auto-save-files's value is t

I have the same setting - it seems to be the default.

>     Note that auto-save file will not be deleted if the buffer is killed
>     when it has unsaved changes.

This is not the behavior I observe.


paul

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

"Matt Armstrong" <matt+dated+1007572794.1eb86e@lickey.com> writes:

> Samuel, could you be thinking of the backup file and not the autosave
> file?

I'm thinking of the autosave file.

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

prj@po.cwru.edu (Paul Jarc) writes:

> Samuel Padgett <spadgett1@nc.rr.com> wrote:
>> C-h v delete-auto-save-files RET:
>>
>>     delete-auto-save-files's value is t
>
> I have the same setting - it seems to be the default.
>
>>     Note that auto-save file will not be deleted if the buffer is killed
>>     when it has unsaved changes.
>
> This is not the behavior I observe.

You're right.  I don't see this either with Emacs 20 or 21.  This is
probably some sort of bug then--maybe a documentation bug?

I will submit a bug report and see what the maintainers say.

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Matt Armstrong]

Samuel Padgett <spadgett1@nc.rr.com> writes:

> "Matt Armstrong" <matt+dated+1007571874.3d8167@lickey.com> writes:
>
>> I think we agree.  C-x k does not delete the file from disk if it has
>> been explicitly saved -- it does get rid of any autosave file and that
>> is fine with me.
>
> C-x k should not delete the autosave file of a modified
> buffer--unless the docstring for `delete-auto-save-files' is wrong.
> Or unless I'm reading it wrong.  Does it for you?
>
> Maybe this is an Emacs bug?

Yes, looks like the documentation doesn't match the behavior.  It
says:

    Note that auto-save file will not be deleted if the buffer is
    killed when it has unsaved changes.

I hope this is a doc bug.  If this were implemented, you'd be prompted
about unsaved changes when killing a buffer, and then again about
recovering the unsaved changes every time you opened the same file
until you deleted the autosave file.  That is just annoying, and has
nothing to do with crash recovery (which is what autosave is
protecting against).

-- 
matt

Re: Is this a buglet.. nndrafts:draft renegades

[Harry Putnam]

Samuel Padgett <spadgett1@nc.rr.com> writes:

> Harry Putnam <reader@newsguy.com> writes:
>
>> Type:
>> Testing the action of C-x k on new composition buffer of tnew
>> after an auto-save has occured
>> <waiting>
>> excerpt from *Nessages*
>>         Auto-saving...done
>>
>> <C-x k yes>
>>
>> C-x d <RET> g 
>> (cong without a trace)
    ^gone
>
> Was the buffer modified when you killed it?

As shown above I visited a file and typed into the buffer as
indicated. Waited for auto-save then C-x k

Re: Is this a buglet.. nndrafts:draft renegades

[Per Abrahamsen]

"Matt Armstrong" <matt+dated+1007576502.e9e0d8@lickey.com> writes:

> Yes, looks like the documentation doesn't match the behavior.  

Remember to send a bug report.

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

Per Abrahamsen <abraham@dina.kvl.dk> writes:

> "Matt Armstrong" <matt+dated+1007576502.e9e0d8@lickey.com> writes:
>
>> Yes, looks like the documentation doesn't match the behavior.  
>
> Remember to send a bug report.

I have.  I got this response from RMS:

    I decided to change the code to fix the documentation.
    Thanks for reporting the bug.

I guess this is a doc bug then?

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Harry Putnam]

Samuel Padgett <spadgett1@nc.rr.com> writes:

> Per Abrahamsen <abraham@dina.kvl.dk> writes:
>
>> "Matt Armstrong" <matt+dated+1007576502.e9e0d8@lickey.com> writes:
>>
>>> Yes, looks like the documentation doesn't match the behavior.  
>>
>> Remember to send a bug report.
>
> I have.  I got this response from RMS:
>
>     I decided to change the code to fix the documentation.
>     Thanks for reporting the bug.
>
> I guess this is a doc bug then?

Hard to tell what is meant there.  Sounds like it means the doc is
right and the behavior wrong.  I hope not..

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

Harry Putnam <reader@newsguy.com> writes:

> Samuel Padgett <spadgett1@nc.rr.com> writes:
>
>> I got this response from RMS:
>>
>>     I decided to change the code to fix the documentation.
>>     Thanks for reporting the bug.
>>
>> I guess this is a doc bug then?
>
> Hard to tell what is meant there.  Sounds like it means the doc is
> right and the behavior wrong.  I hope not..

I asked Richard for clarification, and he said he changed the behavior
to match what the doc says.

Sam

Re: Is this a buglet.. nndrafts:draft renegades

[Harry Putnam]

Samuel Padgett <spadgett1@nc.rr.com> writes:

> Harry Putnam <reader@newsguy.com> writes:
>
>> Samuel Padgett <spadgett1@nc.rr.com> writes:
>>
>>> I got this response from RMS:
>>>
>>>     I decided to change the code to fix the documentation.
>>>     Thanks for reporting the bug.
>>>
>>> I guess this is a doc bug then?
>>
>> Hard to tell what is meant there.  Sounds like it means the doc is
>> right and the behavior wrong.  I hope not..
>
> I asked Richard for clarification, and he said he changed the behavior
> to match what the doc says.

Let me make sure I'm following here.
Existing doc says:

   delete-auto-save-files's value is t

   Documentation:
   *Non-nil means delete auto-save file when a buffer is saved or killed.
   
   Note that auto-save file will not be deleted if the buffer is killed
   when it has unsaved changes.

So the new behavior in future releases will be in accord with that
documentation:

If one `C-x C-f file' when no actual file is on disk, then types into
it. Then aborts with `C-x k <YES>'  An auto-save file will remain on
disk.  That is, piles of #file# will soon develop.  At least with my
usage pattern.

20.7 doesn't do that either.  At least not by default.

Is that your take on RMSs' response?

Re: Is this a buglet.. nndrafts:draft renegades

[Samuel Padgett]

Harry Putnam <reader@newsguy.com> writes:

> If one `C-x C-f file' when no actual file is on disk, then types into
> it. Then aborts with `C-x k <YES>'  An auto-save file will remain on
> disk.  That is, piles of #file# will soon develop.  At least with my
> usage pattern.
>
> 20.7 doesn't do that either.  At least not by default.
>
> Is that your take on RMSs' response?

That is my take on Richard's response.  If you don't think this is
proper behavior, you might want to complain on the emacs-devel list.
Or maybe report a bug.

You also might want to get in the habit of using "M-~ C-x k RET" ;-)

Sam