1021218 - merge-users.pl fails if both users have bug_user_last_visit entries for the same bug

Status: RESOLVED FIXED

(Bugzilla :: Administration, task)

Description

[David Lawrence [:dkl]]

When running a user merge for bug 1021105, the following error occurred:

bugzillaadm.private.scl3# time perl -T ./contrib/merge-users.pl [e-mail] [e-mail]
OK, old user account [e-mail] found; user ID: 447677.
OK, new user account [e-mail] found; user ID: 508500.
.
.
.
OK, records in the 'watched' column of the 'watch' table
have been migrated to the new user account.
DBD::mysql::st execute failed: Duplicate entry '508500-1021105' for key 'bug_user_last_visit_idx' [for Statement "UPDATE bug_user_last_visit
                                           SET user_id = ?
                                         WHERE user_id = ?" with ParamValues: 0=508500, 1=447677] at ./contrib/merge-users.pl line 217

This is due to the unique index on user_id,bug_id and the way merge-users.pl updates user id values in tables that have a foreign key relationship to profiles.userid.

I suppose the solution would be 1) use a query to find any duplicates and 2) use separate queries to remove the row with the old user id and leave the other in place.

Thoughts? 
dkl

Comment 1

[:glob ✱]

merge-users is from upstream, moving.

i suspect the right thing to do is to retain the most recent last-visit time for each bug.

Comment 2

[David Lawrence [:dkl]]

Attachment: 1021218_1.patch

Comment 3

[:glob ✱]

Comment on attachment 8435884 [details] [diff] [review]
1021218_1.patch

Review of attachment 8435884 [details] [diff] [review]:
-----------------------------------------------------------------

r=glob

switch to using earlier.id on commit please.

::: contrib/merge-users.pl
@@ +148,5 @@
> +# Special care needs to be done with bug_user_last_visit table as the
> +# source user and destination user may have visited the same bug id at one time.
> +# In this case we remove the one with the oldest timestamp.
> +my $dupe_bug_entries = $dbh->selectall_arrayref("
> +    SELECT DISTINCT earlier.user_id, earlier.bug_id, earlier.last_visit_ts

it would be better to select earlier.id and delete rows by id

Comment 4

[David Lawrence [:dkl]]

To ssh://[e-mail]/bugzilla/bugzilla.git
   1f859a3..ea10fef  master -> master