Bug in remote file save.

Bug in remote file save.

[Mike Baranski]

Couldn't send this via email, so here it is:

This bug report will be sent to the Free Software Foundation,
not to your local site managers!
Please write in English, because the Emacs maintainers do not have
translators to read other languages for them.

Your bug report will be posted to the bug-gnu-emacs@gnu.org mailing list,
and to the gnu.emacs.bug news group.

In GNU Emacs 21.2.1 (i386-redhat-linux-gnu, X toolkit, Xaw3d scroll bars)
 of 2002-04-08 on porky.devel.redhat.com
configured using `configure  i386-redhat-linux --prefix=/usr 
--exec-prefix=/usr --bindir=/usr/bin --sbindir=/usr/sbin --sysconfdir=/etc 
--datadir=/usr/share --includedir=/usr/include --libdir=/usr/lib 
--libexecdir=/usr/libexec --localstatedir=/var --sharedstatedir=/usr/com 
--mandir=/usr/share/man --infodir=/usr/share/info --with-gcc --with-pop 
--with-sound'
Important settings:
  value of $LC_ALL: nil
  value of $LC_COLLATE: nil
  value of $LC_CTYPE: nil
  value of $LC_MESSAGES: nil
  value of $LC_MONETARY: nil
  value of $LC_NUMERIC: nil
  value of $LC_TIME: nil
  value of $LANG: en_US
  locale-coding-system: iso-latin-1
  default-enable-multibyte-characters: t

Please describe exactly what actions triggered the bug
and the precise symptoms of the bug:

I dired a directory remotely,  C-xd /me@server:/var/www/html/website, open 
and edit a file, it saves the first time correctly.  For each further edit, 
it gives a beep and the message "Keyboard macro terminated by command 
ringing the bell".  Then, emacs thinks that I have an old copy of the file, 
and gives the messages confirming that I want to edit an old version, and 
prompting the next time I save to verify that I want to overwrite a newer 
file.  When the beep comes, the file is not saved properly.


Recent input:

f t p SPC b r o k e n <return> w h e n SPC I SPC <backspace> 
<backspace> <backspace> <backspace> <S-backspace> <S-backspace> 
<S-backspace> W h e n SPC I SPC p o <backspace> <backspace> 
o p e n SPC a SPC r e m o v <backspace> t e SPC f i 
l e , SPC <escape> k <escape> k <down> <down> <down> 
<down> <down> <down> <down> <down> <down> <down> <down> 
<down> <down> <down> f <return> <backspace> C-x C-s 
<return> <return> <backspace> <backspace> C-x C-s <escape> 
x r e p o r t - e m <tab> <return>

Recent messages:
Getting PWD...done
Listing /mike@wsc:/var/www/html/olema/...done
Retrieving c_law.php...done
syntax: ((topmost-intro . 1)), indent: 0
Loading time-stamp...done
Writing c_law.php...done
Wrote c_law.php
syntax: ((topmost-intro . 1)), indent: 0 [2 times]
backward-delete-char-untabify: Beginning of buffer
c_law.php changed on disk; really edit the buffer? (y, n, r or C-h) 
byte-code: Keyboard macro terminated by a command ringing the bell

Re: Bug in remote file save.

[Richard Stallman]

Can you please tell us *exactly* what to type to reproduce this?

    In GNU Emacs 21.2.1 (i386-redhat-linux-gnu, X toolkit, Xaw3d scroll bars)
     of 2002-04-08 on porky.devel.redhat.com
    configured using `configure  i386-redhat-linux-gnu ...
    ...

    Please describe exactly what actions triggered the bug
    and the precise symptoms of the bug:

    I dired a directory remotely,  C-xd /me@server:/var/www/html/website, open 
    and edit a file, it saves the first time correctly.  For each further edit, 
    it gives a beep and the message "Keyboard macro terminated by command 
    ringing the bell".  Then, emacs thinks that I have an old copy of the file, 
    and gives the messages confirming that I want to edit an old version, and 
    prompting the next time I save to verify that I want to overwrite a newer 
    file.  When the beep comes, the file is not saved properly.

Re: Bug in remote file save.

[Alan Shutko]

>     I dired a directory remotely, C-xd
>     /me@server:/var/www/html/website, open and edit a file, it saves
>     the first time correctly.  For each further edit, it gives a
>     beep and the message "Keyboard macro terminated by command
>     ringing the bell".

Could you compare the time on your computer with the time on the
remote host?  Or, immediately after you save a file, look at it in
dired to see the timestamp?

I think I've seen things like this when the clock on the remote host
was significantly ahead of my time (by several minutes).  When I try
to save, I get a backtrace somewhat like this (sorry, this is from
21.2, since that's the only machine with access to a misconfigured
clock right now... I'll try a better backtrace when I get to work).

I don't know why it complains about a keyboard macro, but it does.

Debugger entered--Lisp error: (error "Keyboard macro terminated by a command ri$
  beep()
  byte-code("Æ^X^H\204l^@ÇÈÉ    !\"\210Ê^ZË )\227\211^[\fU\203^^^@Ì\202#^@Í^KÎ\$
  ask-user-about-supersession-threat("/tsreast:/user2/intouch/atstest")
  set-buffer-modified-p(t)
  byte-code("^H^Q\n^SÅ\f!\207" [last-coding-system-used coding-system-used file$
  ange-ftp-write-region(1 18 "/tsreast:/user2/intouch/atstest" nil t)
  apply(ange-ftp-write-region (1 18 "/tsreast:/user2/intouch/atstest" nil t))
  ange-ftp-hook-function(write-region 1 18 "/tsreast:/user2/intouch/atstest" ni$
  write-region(1 18 "/tsreast:/user2/intouch/atstest" nil t "/tsreast:/user2/in$
  basic-save-buffer-2()
  basic-save-buffer-1()
  basic-save-buffer()
  save-buffer(1)
* call-interactively(save-buffer)

-- 
Alan Shutko <ats@acm.org> - In a variety of flavors!
I sold my sole to Satan. He's been floundering around ever since.

Re: Bug in remote file save.

[Kim F. Storm]

Alan Shutko <ats@acm.org> writes:

> I don't know why it complains about a keyboard macro, but it does.

It is the (beep) call in ask-user-about-supersession-threat which
issues that warning.

However, for this to happen, the executing-kbd-macro variable must be
non-nil.  This normally only happens when executing a keyboard macro.

Looking through the code, I only found two places where a package
explicitly sets the executing-kbd-macro variable; both of them in the
`calc' package.

The purpose seems to be a kludge to avoid some form of message from
save-buffer, but at the same time calc seems to install some kind of
save-buffer hook (I haven't tried to understand that part), so if that
hook is called with executing-kbd-macro set to non-nil (or the hook
function sets that variable itself), you will get the keyboard macro
error message if an error occurs during save-buffer via ange-ftp...

So my question is: Were you using (or had loaded) calc when this
happened?


> 
> Debugger entered--Lisp error: (error "Keyboard macro terminated by a command ri$
>   beep()
>   byte-code("Æ^X^H\204l^@ÇÈÉ    !\"\210Ê^ZË )\227\211^[\fU\203^^^@Ì\202#^@Í^KÎ\$
>   ask-user-about-supersession-threat("/tsreast:/user2/intouch/atstest")
>   set-buffer-modified-p(t)
>   byte-code("^H^Q\n^SÅ\f!\207" [last-coding-system-used coding-system-used file$
>   ange-ftp-write-region(1 18 "/tsreast:/user2/intouch/atstest" nil t)
>   apply(ange-ftp-write-region (1 18 "/tsreast:/user2/intouch/atstest" nil t))
>   ange-ftp-hook-function(write-region 1 18 "/tsreast:/user2/intouch/atstest" ni$
>   write-region(1 18 "/tsreast:/user2/intouch/atstest" nil t "/tsreast:/user2/in$
>   basic-save-buffer-2()
>   basic-save-buffer-1()
>   basic-save-buffer()
>   save-buffer(1)
> * call-interactively(save-buffer)
> 

-- 
Kim F. Storm <storm@cua.dk> http://www.cua.dk

Re: Bug in remote file save.

[Alan Shutko]

storm@cua.dk (Kim F. Storm) writes:

> So my question is: Were you using (or had loaded) calc when this
> happened?

No, that machine doesn't have calc anywhere... it happens in Emacs
21.2 with emacs -q --no-site-file.  (I was too busy today to check
the backtrace with CVS, sorry.)  I'll look into it some more....

When this happens, you don't even see the first prompt.  It shouldn't
beep until after it's given you a chance to answer a question... but
I don't see it.

If I edebug-defun, it shows the question, and I can confirm that
executing-kbd-macro is t.  Ah, there it is!  In 21.2,
ange-ftp-write-region binds executing-kbd-macro.  RMS fixed this in
July, but the patch hasn't made its way to the RC branch.

2002-07-12  Richard M. Stallman  <rms@gnu.org>

	* net/ange-ftp.el (ange-ftp-write-region):
	Don't bind executing-kbd-macro; instead, pass non-nil VISIT arg
	to ange-ftp-real-write-region.


-- 
Alan Shutko <ats@acm.org> - In a variety of flavors!
She was only the undertaker's daughter, but any man cadaver.

Re: Bug in remote file save.

[Richard Stallman]

      RMS fixed this in
    July, but the patch hasn't made its way to the RC branch.

    2002-07-12  Richard M. Stallman  <rms@gnu.org>

	    * net/ange-ftp.el (ange-ftp-write-region):
	    Don't bind executing-kbd-macro; instead, pass non-nil VISIT arg
	    to ange-ftp-real-write-region.

I will put that fix into RC, but we still should fix Calc.
Thanks to both of you for tracking this down and reporting the
problem in Calc.

Re: Bug in remote file save.

[Richard Stallman]

    However, for this to happen, the executing-kbd-macro variable must be
    non-nil.  This normally only happens when executing a keyboard macro.

    Looking through the code, I only found two places where a package
    explicitly sets the executing-kbd-macro variable; both of them in the
    `calc' package.

    The purpose seems to be a kludge to avoid some form of message from
    save-buffer, but at the same time calc seems to install some kind of
    save-buffer hook (I haven't tried to understand that part), 

I think it is wrong for any package to set executing-knd-macro except
because it is executing a keyboard macro.  I think Calc needs to be
fixed.

Colin, can you work on fixing this?

Re: Bug in remote file save.

[Mike Baranski]

It was the time issue.  The time was ahead on the remote server by 4 minutes, 
so that fixed it.  Can this option thing be changed?  Some files are on our 
local network, so I can set the machines to have the correct time, but for 
editing really remote files, this can be a problem...  Sorry for the false 
alarm, but thanks for the prompt response.

Mike.

On Thursday 12 September 2002 10:13 pm, you wrote:
> Can you please tell us *exactly* what to type to reproduce this?
>
>     In GNU Emacs 21.2.1 (i386-redhat-linux-gnu, X toolkit, Xaw3d scroll
> bars) of 2002-04-08 on porky.devel.redhat.com
>     configured using `configure  i386-redhat-linux-gnu ...
>     ...
>
>     Please describe exactly what actions triggered the bug
>     and the precise symptoms of the bug:
>
>     I dired a directory remotely,  C-xd /me@server:/var/www/html/website,
> open and edit a file, it saves the first time correctly.  For each further
> edit, it gives a beep and the message "Keyboard macro terminated by command
> ringing the bell".  Then, emacs thinks that I have an old copy of the file,
> and gives the messages confirming that I want to edit an old version, and
> prompting the next time I save to verify that I want to overwrite a newer
> file.  When the beep comes, the file is not saved properly.

-- 
##############################
#          Mike Baranski                   #
# Security Management Consulting  #
#      http://www.secmgmt.com         #
#           919-788-9200                 #
#############################

Re: Bug in remote file save.

[Richard Stallman]

It seems like a bug to me that a Emacs is unable to visit a file
because of a time mismatch.  And it is surely a bug that the error
message talks about keyboard macros.

Alan, do you know anything about why that happens, or where in the
code causes it?

Re: Bug in remote file save.

[Alan Shutko]

Richard Stallman <rms@gnu.org> writes:

> It seems like a bug to me that a Emacs is unable to visit a file
> because of a time mismatch.  And it is surely a bug that the error
> message talks about keyboard macros.

It's not unable to visit the file.  It's unable to save it.  And the
base cause makes sense, which is why I never reported it.

When you save a file to a machine with a clock significantly ahead of
yours, the timestamp on the remote file will be well past the time
Emacs thinks it saved it.  So it'll think it's been modified by some
other program and will check to make sure you really want to save it.

The only reason this doesn't work is the (beep) thinks it's in a
keyboard macro, which you fixed in July on HEAD.  That code should
probably be put in the RC branch, unless it depends on something
else... I didn't trace it too far.

-- 
Alan Shutko <ats@acm.org> - In a variety of flavors!
Start showing up early for work