Open Bug 597329 Opened 14 years ago Updated 3 years ago

prefs.js reset (all settings lost) should a network failure occur on profile's network share

Categories

(MailNews Core :: Networking, defect)

x86
Windows XP
defect
Not set
critical

Tracking

(Not tracked)

UNCONFIRMED

People

(Reporter: marc3, Unassigned)

References

Details

(Keywords: dataloss)

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 6.1; fr; rv:1.9.2.4) Gecko/20100611 Firefox/3.6.4
Build Identifier: 3.1.3

The prefs.js file is reset (all settings lost) should a network failure occur on an offline folder on which the profile is located.

Reproducible: Always

Steps to Reproduce:
1. Store your profile on a network share (eg H:\mail\thunderbird), which is synchronized and declared as offline capable.
2. Disconnect the server => the share becomes offline
3. Stop / start thunderbird and enjoy your new almost blank prefs.js !



There was no such problem with TB 2...
Severity: normal → critical
Keywords: dataloss
Summary: prefs.js reset (all settings lost) should a network failure occur on an offline folder → prefs.js reset (all settings lost) should a network failure occur on profile's network share
Ouch. This is another symptom of some code not checking
the failure basic low-level I/O, which I noticed for pop3 and imap downloading in TB.
But I thought pref.js ought to use, "safe-stream" or whatever is more reliable than
ordinary file I/O. 
Maybe "safe" is not that dependable :-(
Component: General → Networking
Product: Thunderbird → MailNews Core
See Also: → 905576
I'd guess prefs.js file is written by code in mozilla-central, not any of out networking code.

I have a very similar issue with TB 78.10.0 (64-Bit) on Linux. The user profiles reside on network shares on a samba file server. This works in 9 out of 10 tries or so but from time to time the prefs.js get reset and therefore all settings are lost. Prefs.js has a size of 0 byte then. Restoring the prefs.js from a backup will get the profile working again. But the difference to the original report is that the share is not unmounted at any time. The damage happens on program close. It is not necessary to shut the machine down, simply closing TB is enough to trigger the failure. Again - this only happens from time to time. I suppose it is triggered by network latency or some other thing causing slow storing speed.

You need to log in before you can comment on or make changes to this bug.