GNU bug report logs - #23618
Some filenotify tests fail with extra "changed" events

Previous Next

To add a comment to this bug, you must first unarchive it, by sending
a message to control AT debbugs.gnu.org, with unarchive 23618 in the body.
You can then email your comments to 23618 AT debbugs.gnu.org in the normal way.

Toggle the display of automated, internal messages from the tracker.

Message #3 received at submit <at> debbugs.gnu.org (full text, mbox):

From: Glenn Morris <rgm <at> gnu.org>
To: submit <at> debbugs.gnu.org
Subject: Some filenotify tests fail with extra "changed" events
Date: Wed, 25 May 2016 12:55:44 -0400

[Message part 1 (text/plain, inline)]

Package: emacs
Version: 25.0.94
Severity: minor

Some filenotify tests have failed for some time (always?) for me on RHEL 7.2.
Finally a report about it. In the past I've only ever seen one extra
"changed", never the silly number the following log includes! :)

PS as you can see from the log, these tests took ~ 15 minutes to run.

[file-notify-tests.log.gz (application/octet-stream, attachment)]

Message #6 received at 23618 <at> debbugs.gnu.org (full text, mbox):

From: Michael Albinus <michael.albinus <at> gmx.de>
To: Glenn Morris <rgm <at> gnu.org>
Cc: 23618 <at> debbugs.gnu.org
Subject: Re: bug#23618: Some filenotify tests fail with extra "changed" events
Date: Fri, 27 May 2016 11:24:01 +0200

Glenn Morris <rgm <at> gnu.org> writes:

Hi Glenn,

> Package: emacs
> Version: 25.0.94
> Severity: minor
>
> Some filenotify tests have failed for some time (always?) for me on RHEL 7.2.
> Finally a report about it. In the past I've only ever seen one extra
> "changed", never the silly number the following log includes! :)

Hard to say something about without debugging. Due to the nature of file
notification events, those `changed' events could appear from anywhere.

As you might imagine, I cannot reproduce the problems locally. Also
hydra does not report about.

The only exception is `file-notify-test08-watched-file-in-watched-dir-remote',
which has failed here as well. The remote `gvfs-monitor-dir' library is
a little bit fickle.

We have done some work in the master branch, which is not contained in
the emacs-25 branch. I don't see any problem running the tests in the
master branch. Do you see the problems also in the master branch?

It might be also interesting to know whether the failed tests appear
also during an ordinary "make check". I suppose you have applied "make
-C test/automated file-notify-tests", which runs the expensive tests.

> PS as you can see from the log, these tests took ~ 15 minutes to run.

Yes. The expensive tests contain tests with a looooot of events, because
some of the problems have appeared only in stress situations. On my
6-years-old laptop, running Ubuntu 16.04, the tests take ~12 minutes.

Best regards, Michael.

Message #9 received at 23618 <at> debbugs.gnu.org (full text, mbox):

From: Glenn Morris <rgm <at> gnu.org>
To: Michael Albinus <michael.albinus <at> gmx.de>
Cc: 23618 <at> debbugs.gnu.org
Subject: Re: bug#23618: Some filenotify tests fail with extra "changed" events
Date: Fri, 27 May 2016 12:38:03 -0400

Michael Albinus wrote:

> We have done some work in the master branch, which is not contained in
> the emacs-25 branch. I don't see any problem running the tests in the
> master branch. Do you see the problems also in the master branch?

I do.

> It might be also interesting to know whether the failed tests appear
> also during an ordinary "make check".

Current master make check:

3 unexpected results:
   FAILED  file-notify-test02-events
   FAILED  file-notify-test04-file-validity
   FAILED  file-notify-test07-backup

Still due to extra "changed" events.

Message #12 received at 23618 <at> debbugs.gnu.org (full text, mbox):

From: Michael Albinus <michael.albinus <at> gmx.de>
To: Glenn Morris <rgm <at> gnu.org>
Cc: 23618 <at> debbugs.gnu.org
Subject: Re: bug#23618: Some filenotify tests fail with extra "changed" events
Date: Fri, 27 May 2016 20:00:53 +0200

Glenn Morris <rgm <at> gnu.org> writes:

>> We have done some work in the master branch, which is not contained in
>> the emacs-25 branch. I don't see any problem running the tests in the
>> master branch. Do you see the problems also in the master branch?
>
> I do.
>
>> It might be also interesting to know whether the failed tests appear
>> also during an ordinary "make check".
>
> Current master make check:
>
> 3 unexpected results:
>    FAILED  file-notify-test02-events
>    FAILED  file-notify-test04-file-validity
>    FAILED  file-notify-test07-backup
>
> Still due to extra "changed" events.

OK, I'll try to get an RHEL 7.2 VM in order to reproduce it. Or maybe a
CentOS VM, in case RHEL VMs are not available.

In case it fails, I will send you an instrumentation patch for
filenotify-tests.el. We do agree that we work on the master branch,
don't we?

Best regards, Michael.

Message #15 received at 23618 <at> debbugs.gnu.org (full text, mbox):

From: Michael Albinus <michael.albinus <at> gmx.de>
To: Glenn Morris <rgm <at> gnu.org>
Cc: 23618 <at> debbugs.gnu.org
Subject: Re: bug#23618: Some filenotify tests fail with extra "changed" events
Date: Sun, 05 Jun 2016 19:59:11 +0200

Michael Albinus <michael.albinus <at> gmx.de> writes:

Hi Glenn,

> OK, I'll try to get an RHEL 7.2 VM in order to reproduce it. Or maybe a
> CentOS VM, in case RHEL VMs are not available.

I haven't been able to reproduce the problems this way :-(

Anyway, the bug in `file-notify-test08-watched-file-in-watched-dir-remote'
I could fix (this was the only one I could reproduce locally).

Furthermore, I've added code to filenotify-tests.el in order to make it
more robust. That' means especially, that no events could arrive during
a test, resulting from a *previous* test. Maybe this fixes the other
problems already. Committed to the master branch.

> In case it fails, I will send you an instrumentation patch for
> filenotify-tests.el. We do agree that we work on the master branch,
> don't we?

In that case, pls apply the following patch

--8<---------------cut here---------------start------------->8---
*** /home/albinus/src/emacs/lisp/filenotify.el.~6627dd7c9405670db99b792f782aa94c5da118a9~	2016-06-05 19:56:18.485624177 +0200
--- /home/albinus/src/emacs/lisp/filenotify.el	2016-06-05 19:56:02.353342656 +0200
***************
*** 86,92 ****
  
  Otherwise, signal a `file-notify-error'."
    (interactive "e")
!   ;;(message "file-notify-handle-event %S" event)
    (if (and (eq (car event) 'file-notify)
  	   (>= (length event) 3))
        (funcall (nth 2 event) (nth 1 event))
--- 86,92 ----
  
  Otherwise, signal a `file-notify-error'."
    (interactive "e")
!   (message "file-notify-handle-event %S" event)
    (if (and (eq (car event) 'file-notify)
  	   (>= (length event) 3))
        (funcall (nth 2 event) (nth 1 event))
***************
*** 261,270 ****
  		    (and (stringp file1)
  			 (string-equal
  			  (nth 0 entry) (file-name-nondirectory file1)))))
!           ;;(message
!            ;;"file-notify-callback %S %S %S %S %S"
!            ;;(file-notify--descriptor desc (car entry))
!            ;;action file file1 registered)
  	  (if file1
  	      (funcall
  	       callback
--- 261,270 ----
  		    (and (stringp file1)
  			 (string-equal
  			  (nth 0 entry) (file-name-nondirectory file1)))))
!           (message
!            "file-notify-callback %S %S %S %S %S"
!            (file-notify--descriptor desc (car entry))
!            action file file1 registered)
  	  (if file1
  	      (funcall
  	       callback
--8<---------------cut here---------------end--------------->8---

Hopefully, the logs will tell us what happens.

Best regards, Michael.

Message #18 received at 23618 <at> debbugs.gnu.org (full text, mbox):

From: Glenn Morris <rgm <at> gnu.org>
To: Michael Albinus <michael.albinus <at> gmx.de>
Cc: 23618 <at> debbugs.gnu.org
Subject: Re: bug#23618: Some filenotify tests fail with extra "changed" events
Date: Tue, 07 Jun 2016 17:23:22 -0400

[Message part 1 (text/plain, inline)]

Please see attached.

[filenotify-tests.log.xz (application/octet-stream, attachment)]

Message #21 received at 23618 <at> debbugs.gnu.org (full text, mbox):

From: Michael Albinus <michael.albinus <at> gmx.de>
To: Glenn Morris <rgm <at> gnu.org>
Cc: 23618 <at> debbugs.gnu.org
Subject: Re: bug#23618: Some filenotify tests fail with extra "changed" events
Date: Mon, 20 Jun 2016 14:46:33 +0200

Hi Glenn,

I've stared on the logs for several days. Still unsure what happened. So
I've weakened the tests a little bit. Let's see how it works.

Thanks, and best regards, Michael.

Message #24 received at 23618 <at> debbugs.gnu.org (full text, mbox):

From: Glenn Morris <rgm <at> gnu.org>
To: Michael Albinus <michael.albinus <at> gmx.de>
Cc: 23618 <at> debbugs.gnu.org
Subject: Re: bug#23618: Some filenotify tests fail with extra "changed" events
Date: Thu, 23 Jun 2016 13:34:39 -0400

[Message part 1 (text/plain, inline)]

It's better, thanks.
file-notify-test08-watched-file-in-watched-dir-remote still fails, for
impenetrable reasons. But it isn't run by default, so maybe it's time to
stop thinking about this now. The expensive tests are so slow, no-one
but you will ever run them anyway. ;)

[filenotify-tests.log.xz (application/octet-stream, attachment)]

Message #29 received at 23618-done <at> debbugs.gnu.org (full text, mbox):

From: Michael Albinus <michael.albinus <at> gmx.de>
To: Glenn Morris <rgm <at> gnu.org>
Cc: 23618-done <at> debbugs.gnu.org
Subject: Re: bug#23618: Some filenotify tests fail with extra "changed" events
Date: Thu, 23 Jun 2016 20:34:13 +0200

Glenn Morris <rgm <at> gnu.org> writes:

> It's better, thanks.

Puh, finally ...

> file-notify-test08-watched-file-in-watched-dir-remote still fails, for
> impenetrable reasons. But it isn't run by default, so maybe it's time to
> stop thinking about this now. The expensive tests are so slow, no-one
> but you will ever run them anyway. ;)

It's not that bad, all created events have passed, and most of the
changed events. In reality I don't expect any problem, because changed
events appear more than once for a given change.

As you could assume, I have no problems locally. I guess it is a problem
of timing, but I don't want to increase the delays longer than they are.

Yes, let's close this bug. Thanks for your patient testing!

Best regards, Michael.

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.