bug#12863: Annoying bug in revert-buffer

bug#12863: Annoying bug in revert-buffer

[Glenn Morris]

Cecil Westerhof wrote:

> I use revert-buffer on several log buffers. There is one annoying bug.
> Sometimes Emacs sees that the buffer has changed and reloads it. But
> when it changes in the same second again, the buffer is not reloaded
> and you do not see the changes. First I thought it was a Windows
> quirk, but now I have seen the same behaviour on Linux. (I am looking
> more at log buffers I suppose.)

Are you talking about auto-revert-mode?
If so, C-h v auto-revert-interval (default 5 seconds).

Otherwise, please tell us which version of Emacs this is.

bug#12863: Annoying bug in revert-buffer

[Glenn Morris]

Glenn Morris wrote:

> Are you talking about auto-revert-mode?
> If so, C-h v auto-revert-interval (default 5 seconds).

Actually that's probably irrelevant.

bug#12863: Annoying bug in revert-buffer

[Glenn Morris]

(Please use reply-to-all)

Cecil Westerhof wrote (on Sun, 11 Nov 2012 at 21:25 +0100):

> >> Are you talking about auto-revert-mode?
> >> If so, C-h v auto-revert-interval (default 5 seconds).
> >
> > Actually that's probably irrelevant.
> 
> Yes that is where I am talking about. Emacs sees that the file is
> changed and reloads it in the buffer. But if in the same second the
> file changes again, it is not reloaded.
> 
> Here I am using:
>     GNU Emacs 23.3.1 (i586-suse-linux-gnu, GTK+ Version 2.24.7) of 2011-10-30 on build34
> 
> But I have seen the same problem under Windows. At the moment I do not
> know which version that was. If needed I can retrieve that one also.
> 
> A solution would be: (also) reloading a second after a change.

I'm still unsure if you are talking about auto-revert-mode or an
explicit call to revert-buffer. If I read the code correctly,
revert-buffer does not care about time-stamps. auto-revert-mode does,
and I think in Emacs 24.2 and earlier, file time-stamps were only
stored to second precision. So indeed in Emacs 24.2 and earlier, file
modifications that happened within a second of reverting would
probably be invisible to auto-revert mode.

Emacs 24.3 (not yet released) will use much higher resolution
time-stamps, so this issue should be fixed there.

bug#12863: Annoying bug in revert-buffer

[Cecil Westerhof]

Glenn,

Op maandag 12 nov 2012 02:07 CET schreef Glenn Morris:

>
> (Please use reply-to-all)
>
> Cecil Westerhof wrote (on Sun, 11 Nov 2012 at 21:25 +0100):
>
>>>> Are you talking about auto-revert-mode?
>>>> If so, C-h v auto-revert-interval (default 5 seconds).
>>>
>>> Actually that's probably irrelevant.
>>
>> Yes that is where I am talking about. Emacs sees that the file is
>> changed and reloads it in the buffer. But if in the same second the
>> file changes again, it is not reloaded.
>>
>> Here I am using:
>> GNU Emacs 23.3.1 (i586-suse-linux-gnu, GTK+ Version 2.24.7) of 2011-10-30 on
>> 	build34
>>
>> But I have seen the same problem under Windows. At the moment I do not
>> know which version that was. If needed I can retrieve that one also.
>>
>> A solution would be: (also) reloading a second after a change.
>
> I'm still unsure if you are talking about auto-revert-mode or an
> explicit call to revert-buffer.

I need to communicate better. :-{

It is auto-revert-mode. When I see that it has gone wrong, I can
correct it with revert-buffer. ;-}


> If I read the code correctly,
> revert-buffer does not care about time-stamps. auto-revert-mode does,
> and I think in Emacs 24.2 and earlier, file time-stamps were only
> stored to second precision. So indeed in Emacs 24.2 and earlier, file
> modifications that happened within a second of reverting would
> probably be invisible to auto-revert mode.

But if you would revert the buffer (also) in the second after the
change, the problem would be solved,


> Emacs 24.3 (not yet released) will use much higher resolution
> time-stamps, so this issue should be fixed there.

Any idea when it will come out?

OpenSuse are not very good with maintaining Emacs. My version is 3½
years old and there have been five new releases. Maybe I should
install Emacs manually.

-- 
Cecil Westerhof
Senior Software Engineer
M Cecil@Decebal.nl

LinkedIn: http://www.linkedin.com/in/cecilwesterhof

O< ascii ribbon campaign - stop html mail - www.asciiribbon.org

Please do not send me Microsoft Office/Apple iWork documents.
Send OpenDocument instead! http://fsf.org/campaigns/opendocument/

bug#12863: Annoying bug in revert-buffer

[Glenn Morris]

Actually, maybe this isn't fixed. I'm having a hard time testing it, but
I see this comment in verify-visited-file-modtime:

/* If both [times] exist, accept them if they are off by one second.  */

Cc'ing Paul since he has worked on the time-stamps issue and can
probably tell us whether or not it is fixed...

bug#12863: Annoying bug in revert-buffer

[Andreas Schwab]

Glenn Morris <rgm@gnu.org> writes:

> Emacs 24.3 (not yet released) will use much higher resolution
> time-stamps, so this issue should be fixed there.

Only if the filesystem supports sub-second resolution.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

bug#12863: Annoying bug in revert-buffer

[Stefan Monnier]

> Emacs 24.3 (not yet released) will use much higher resolution
> time-stamps, so this issue should be fixed there.

Maybe we should also pay attention to the file-size.


        Stefan

bug#12863: Annoying bug in revert-buffer

[Paul Eggert]

On 11/12/2012 12:04 PM, Glenn Morris wrote:
> I see this comment in verify-visited-file-modtime:
> 
> /* If both [times] exist, accept them if they are off by one second.  */

I've had my eye on that code for some time, figuring that
it must be bogus but not having the time to prove that it
was bogus.  Since it's now known to cause problems, I removed it
in trunk bzr 110875.  Since this is not a recent regression
I did not install the change into the emacs-24 branch.

bug#12863: Annoying bug in revert-buffer

[Glenn Morris]

Version: 24.4

Paul Eggert wrote:

> On 11/12/2012 12:04 PM, Glenn Morris wrote:
>> I see this comment in verify-visited-file-modtime:
>> 
>> /* If both [times] exist, accept them if they are off by one second.  */
>
> I've had my eye on that code for some time, figuring that
> it must be bogus but not having the time to prove that it
> was bogus.  Since it's now known to cause problems, I removed it
> in trunk bzr 110875. 

Thanks, I checked using `touch' and verify-visited-file-modtime that it
fixes the issue for me.

So this will be fixed in the next (probably) release after 24.3, not in
24.3 as I originally said.