462006 - prefs.js get truncated.

Status: RESOLVED INCOMPLETE

(Thunderbird :: Preferences, defect)

Description

[Roger Gammans]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686 (x86_64); en-US; rv:1.8.1.16) Gecko/20080702 Iceweasel/2.0.0.16 (Debian-2.0.0.16-0etch1)
Build Identifier: 2.0.0.17 (20080914)

After using thunderbird a couple of time the next time it is open there are no mail folders showing. Invetsigating shows prefs.js is truncated and finishes in the middle of a line.

Reproducible: Sometimes

Steps to Reproduce:
1. Open thunderbird 
2. setup  an imap accounts. (Branded google mail from sky.com as an Imap account)
3. Read mail.
4. close thunderbird
5  open thunderbird
6. repeat from 3.

I'm sorry I can't be more specific as I can't reproduce accurately enough but it happens on the users PC about 2 a week.
Actual Results:  
When Thunderbird is open the profile setup wizard appears as has now account details defined in the prefs.js file.

Expected Results:  
The mail folders should still be showing, and the prefs.js .

This happens at random - presuambly on saving the prefs.js file.

I couldn't see any backup files in the profile directory - but I'm not entirely sure what they would be named.

The machine has over 250GB of disk space free so it not a disk space issue unless there is an overflow somewhere.

Comment 1

[Matthias Versen [:Matti]]

It could happen if you don't close Thunderbird manually before shutting down windows and windows shuts down and powers off the system faster than the HDD could write it's SRAM Cache to the disk itself.

If you can find any reasons for this then reopen this bug and we will try to fix that but without more information we can only dupe this to the tracking bug.

Comment 2

[Matthias Versen [:Matti]]

see also bug 415910, bug 304045

Comment 3

[WADA:World Anti-bad-Duping Agency]

To Matthias Versen:
Bug 193638 is meta bug for tracking of bugs for specific problem. DUPing of bug for specific problem to meta bug(for tracking purpose) is incorrect action.
Please note that "closing as INCOMPLETE" is lately used for bug of insufficient/incomplete data and no response from bug opener for long time.

Re-opening, with setting dependency of Bug 193638 (193638 in Blocks: of this bug).

Comment 4

[Roger Gammans]

Matthias - thats a nice theory,  but do you have any evidence?

I'll speak to the user and inform him of this anyway , and switch off rite-caching - but this implies a windows bug which isn't that likely. 
( Windows should flush the drive before switching off ). And Kb article http://support.microsoft.com/kb/332023 is interesting in light of this.
- 

Does Thunderbird call fsync() - or the windows analague FlushFileBuffers() [ 
http://msdn.microsoft.com/en-us/library/aa364439(VS.85).aspx ] after writing this file?.

Also bug 193638 mentions prefs.js.bak ad 30405 mentions prefs-n.js these files were not in the profile .

Do you have any other ideas of tests, I can do to help you guys track this down?

Comment 5

[Matthias Versen [:Matti]]

To Wada:
I know what bug 193638 is, I'm the reporter of it.
The lost prefs.js is known and nobody could figure out a reason or do you have an idea what could be the reason for this or figure it out in this case ?
Is your plan to leave 40 unconfirmed bugs in bugzilla without without having more information ?
I don't think it's possible to figure out what's wrong unless a developer can reproduce this on his system.
The plan was to dupe such bugs to 193638 and create new bugs and mark them depend if you find a single reason for such a lost prefs.js.
I can of course use incomplete as a response to the bug opening if I edon't xpect to get more information from the reporter but why do you mention that in this case ?

Roger:
The lost prefs.js is very long known and the developers did a few things to avoid it but it's still happening. I have no idea how to figure out the reason. A part of the problem could be a slow thunderbird on shutdown and windows kills it or whatever. I use Seamonkey since 8 years and i have never lost my prefs.js on different systems but some people hit this problem.
I don't know if thunderbird still writes the prefs.js backup files or if you have to set a hidden pref.

Comment 6

[Wayne Mery (:wsmwk)]

(why hasn't this been marked critical????)

Roger, if comment 0 is true, "happens on the users PC about 2 a week", then your  friend is potentially an excellent testcase. Can you get the person directly involved in this bug?

Also, there is no mention of PC having failed or been restarted leading up to this problem. Can you clarify that no PC hang, restart or other failure is involved?

Comment 7

[WADA:World Anti-bad-Duping Agency]

(Off-Topic)

(In reply to comment #5)
> Is your plan to leave 40 unconfirmed bugs in bugzilla without having more information ?

To Matthias Versen: (sorry for late response)

As I wrote in bug 422447 comment #14, new culprit is being found. And as seen in bug 477934, improvement is being done. If bug is set in dependency tree for meta bug, and if bug is kept open, we can ask bug opener for new information. It's one of reasons why I ask you not to DUP bugs to meta bug.
I think closing as "INCOMPLETE" is appropriate action, if no new information such as specific condition is not provided by bug opener or comment posters for long time. (Bug can be reached via "Show Resolved" of dependency tree.)

Comment 8

[Roger Gammans]

(In reply to comment #6)
> (why hasn't this been marked critical????)
> 
> Roger, if comment 0 is true, "happens on the users PC about 2 a week", then
> your  friend is potentially an excellent testcase. Can you get the person
> directly involved in this bug?

(I wrote a batch file to recopy a prefs.js from a backup ) So the user has been running fine, but I spoke to them again and apparently it has settled down and doesn't reocuur as often as it did to start with . It no occurs 'quite infrequently'

So it's not going to make a good testcase as I'd hoped it would when I wrote comment #0.

Comment 9

[Wayne Mery (:wsmwk)]

Roger, thanks for the update. Please clarify the following:
* quantify, roughly how often is "infrequently"?  
* startup symptom - they get account setup wizard?
* restart - do they simply close Thunderbird and restart, with no intervening PC hang, PC restart or other failure?
* shutdown - is it via file|close/file|exit, or red X, or windows' shutdown?

I am quite in agreement with wada's comment 7. Conflating reports like this into a single bug is not good unless/until the symptoms and the steps are rock solid duplicate of another bug - it is far easier to aggressively QA a single issue and single reporter, if it's a reproducible testcase, than deal with multiple reporters/voices all at once in one big bug. 

It is SO important to mark bugs with critical/dataloss symptoms as criticial/dataloss immediately (like this bug) - even if the report is duped or closed, in case it gets reopened; or in the case of not yet confirmed, so we can get more eyes on it and ramp up the QA.  We don't want to miss good opportunities - reproducible testcases + willing reporter are rare.

Comment 10

[WADA:World Anti-bad-Duping Agency]

As I wrote in bug 422447 comment #14, a new suspect has been found.
Because your case is "truncated prefs.js file" even when normal termination, and because frequency seems very low, I'm suspecting Bug 480466.
Roger Gammans, can you request your user to keep backup of truncated prefs.js when user encounters problem again? If possible, please attach truncated prefs.js file, to see at where truncation occurs.

Comment 11

[[:Aureliano Buendía]]

Roger could answer to comment #9, please?

Comment 12

[Wayne Mery (:wsmwk)]

Roger in comment #8
> (In reply to comment #6)
> > (why hasn't this been marked critical????)
> > 
> > Roger, if comment 0 is true, "happens on the users PC about 2 a week", then
> > your  friend is potentially an excellent testcase. Can you get the person
> > directly involved in this bug?
> 
> (I wrote a batch file to recopy a prefs.js from a backup ) So the user has been
> running fine, but I spoke to them again and apparently it has settled down and
> doesn't reocuur as often as it did to start with . It no occurs 'quite
> infrequently'

mailtweak extension supposedly will do this http://mailtweak.mozdev.org/tweaks.html#prefcache

unfortunately we don't have reproducible steps here, so closing incomplete.
but any additional information from Roger would be appreciated.