Bug#399016: kaddressbook: Data loss, involving strange file duplication and locking problems
On Fri, Nov 17, 2006 at 12:26:13AM -0600, Adam Porter wrote:
> Package: kaddressbook
> Version: 4:3.5.5.dfsg.1-1
> Severity: grave
> Hash: SHA1
> I have noticed this behavior for some time, since back in KDE 3.4.x at least, but this is the last straw, so I'm
> filing a bug. A week or so ago I entered a few new contacts into my KDE address book. When I went to look up the contacts' e-mail addresses just
> now, my search in KABC found nothing. I opened a shell and found this:
> $ cd .kde/share/apps/kabc/
> $ ll
> total 504K
> drwx------ 2 first first 4.0K 2006-11-16 05:15 lock
> drwx------ 2 first first 4.0K 2005-12-19 18:58 logos
> drwx------ 2 first first 4.0K 2005-12-19 18:58 photos
> drwx------ 2 first first 4.0K 2005-12-19 18:58 sounds
> - -rwx------ 1 first first 41K 2006-11-16 05:15 std.vcf
> - -rwx------ 1 first first 41K 2006-11-16 05:15 std.vcf__0
> - -rw-r--r-- 1 first first 41K 2006-11-16 05:15 std.vcf__1
> - -rw-r--r-- 1 first first 41K 2006-11-16 05:15 std.vcf__2
> - -rwx------ 1 first first 0 2006-11-09 03:03 std.vcf2WPY5b.new
> - -rw-r--r-- 1 first first 41K 2006-11-16 05:15 std.vcf__3
> - -rw-r--r-- 1 first first 41K 2006-11-16 05:15 std.vcf__4
> - -rw-r--r-- 1 first first 41K 2006-11-16 05:15 std.vcf__5
> - -rw-r--r-- 1 first first 42K 2006-11-16 05:15 std.vcf__6
> - -rw-r--r-- 1 first first 42K 2006-11-16 05:15 std.vcf__7
> - -rw-r--r-- 1 first first 42K 2006-11-16 05:15 std.vcf__8
> - -rw-r--r-- 1 first first 41K 2006-11-16 05:15 std.vcf__9
> - -rwx------ 1 first first 0 2006-11-13 02:55 std.vcfVRcdrb.new
> $ ll lock
> total 8.0K
> - -rw-r--r-- 1 first first 11 2006-11-09 03:03 _home_first_.kde_share_apps_kabc_std.vcfMuo9TRh0
> - -rw-r--r-- 1 first first 10 2006-11-13 02:55 _home_first_.kde_share_apps_kabc_std.vcfSj652hp5
> $ rgrep lash .
> As you can see, there are many duplicates of the std.vcf file, and two lock files. The dates on the lock files
> don't match the date on the current std.vcf file. There are two old std.vcf...new files that are empty. There
> are three std.vcf__# files that are larger than the rest, and inside those three files are the contacts that I
> entered that are now missing in std.vcf.
> Thankfully, I can copy one of those three files over the std.vcf file and regain those contacts. But what in the
> world is going on here?
> I'm rating this as grave because I consider it a data-loss bug. Most users won't have any idea how to dig for
> the vcf files, and even if they do, there's no guarantee that the missing data will be there. And even worse, if
> this went unnoticed for a while, you could end up with a branching scenario where some files had some contacts,
> and other files had other contacts, and somehow you'd have to merge them back together into a single file to get
> them all back. Worst of all, this makes the KDE address book system (which is used by Kontact/KMail/etc)
> completely unreliable and untrustworthy. I hope Etch isn't released with this bug.
I have been playing with kaddresbook since you reported this bug. I have added,
removed and edited contacts without problems in several testing sessions.
I also have looked to my files below .kde/share/apps/kabc, and the structure
is very similar to yours, so i do not think this is the problem (but
i agree it is confusing).
Could you give more information about this?
> Please let me know how I can help debug this.
Could you try KAddressBook in a different computer and test if you have
the same problems? And in the same computer as a diferent user?