54566 - Software error generated when adding invalid dependencies

Status: VERIFIED FIXED

(Bugzilla :: Bugzilla-General, defect, P3)

Description

[Adam Spiers]

If you try to make a bug depend on or block an invalid bug id, bugzilla
correctly reports in big bad red and black that the bug id you entered
was invalid.  However, it then generates a software error.  I changed
the die() to a confess() and got this stack trace:

SELECT profiles.userid, profiles.groupset, profiles.login_name,
profiles.login_name = '[e-mail]' AND profiles.cryptpassword =
logincookies.cryptpassword AND logincookies.hostname = '192.168.1.1',
profiles.disabledtext  FROM profiles, logincookies WHERE logincookies.cookie =
'310' AND profiles.userid = logincookies.userid: Table 'logincookies' was not
locked with LOCK TABLES at globals.pl line 140
        main::SendSQL('SELECT profiles.userid, profiles.groupset,
profiles.login_name, ...') called at CGI.pl line 538
        main::quietly_check_login() called at CGI.pl line 909
        main::GetCommandMenu() called at globals.pl line 944
        main::Param('commandmenu') called at globals.pl line 977
        main::PerformSubsts('<TABLE BORDER="0"><TR><TD
BGCOLOR="#000000" VALIGN="TOP">^J<TABLE...') called at
CGI.pl line 812
        main::PutFooter() called at CGI.pl line 821
        main::PuntTryAgain('12345 is not a legal bug number') called at
/nfs-home/stuff/web/bugzilla/process_bug.cgi line 727

Why MySQL was expecting logincookies to be locked, I have no idea ...

Comment 1

[Matthew Tuck [:CodeMachine]]

*** Bug 51050 has been marked as a duplicate of this bug. ***

Comment 2

[Dave Miller [:justdave]]

ping cyeh...

Is this the same thing you just fixed in editusers.cgi?  Same problem perhaps?

Comment 3

[Chris Yeh]

It appears to be a similar problem: we lock the tables to begin bug processing 
but when an error occurs we punt without unlocking the tables first.

We're going to have to walk through all the logic of process_bug.cgi and see 
where we should be unlocking the tables in a failure case.

Comment 4

[Chris Yeh]

re-assigning to me, have patch

Comment 5

[Chris Yeh]

Attachment: if we PuntTryAgain, unlock all tables

Comment 6

[Chris Yeh]

checked patch into trunk

Comment 7

[Dave Miller [:justdave]]

In search of accurate queries....  (sorry for the spam)

Comment 8

[Gervase Markham [:gerv]]

VERIFIED. Now works as it should.

Gerv

Comment 9

[Dave Miller [:justdave]]

Moving closed bugs to Bugzilla product