Great Circle Associates Majordomo-Users
(October 1993)
 

Indexed By Date: [Previous] [Next] Indexed By Thread: [Previous] [Next]

Subject: Re: multiple unsubscribes
From: Kingsley Kerce <kerce @ dirac . scri . fsu . edu>
Date: Wed, 13 Oct 1993 09:18:00 -0400
To: Kasuga Jingoro <jingoro @ rahul . net>
Cc: Majordomo-Users @ GreatCircle . COM
In-reply-to: <199310131252.AA11776@bolero.rahul.net>
References: <199310111500.AA96118@dirac.scri.fsu.edu><199310131252.AA11776@bolero.rahul.net>

The following fix works for us.

Thanks, Kasuga (or is it Jim?)

Kingsley F. Kerce (kerce@scri.fsu.edu)
Supercomputer Computations Research Institute
Florida State University
Tallahassee, FL, USA  32306

Kasuga Jingoro writes:
 > > "Brent Chapman" <Brent@GreatCircle.COM> writes:
 > > "Stephen Bannasch" <stephen_bannasch@terc.edu> writes:
 > >  > #           Subject:   multiple unsubscribes
 > >  > # I am the owner of a list and wanted to remove a bunch a bad addresses
 > >  > # so I tried the following commands:
 > >  > # 
 > >  > # approve PASSWORD unsubscribe MYLIST USER1
 > >  > # approve PASSWORD unsubscribe MYLIST USER2
 > >  > # approve PASSWORD unsubscribe MYLIST USER3
 > >  > # 
 > >  > # Majordomo responded by only unsubcribing the first person on the
 > >  > # list.
 > >  > # The second and third were reported as not being members of MYLIST. 
 > >  > # When
 > >  > # the unsubscribe requests for users 2 and 3 were placed in their own 
 > >  > # messages the unsubscribe operations were successful.
 > >  > # 
 > >  > # Is this a bug?  Is there a fix?
 > 
 > This might be related to a related bug in Majordomo.  This is a bug that
 > was reported on this list ages ago but has never been fixed in the
 > official release.  In the do_unsubscribe subroutine, the file NEW is
 > opened to write out the new version of the subscriber list.  But this
 > file is NEVER closed.  The result is that the new version of the list
 > is held open AND unlocked for the duration of the process.  This can cause
 > numerous problems, one of which may be the one you are experiencing.
 > The problem I saw was that if one would place a unsub command followed
 > by a sub command in the same message, the sub command would appear to
 > work, but a later inspection of the list would show that the address
 > was not added.  My theory is that since the new version of the list is
 > held open during the subscribe command, that the changes made to the
 > file in the subscribe command will be superseded once the process ends
 > since perl will then automatically close the NEW file which will revert
 > the file back to the original contents after the unsub command.  Although
 > this bug does not manifest itself on Brent's system, this bug is
 > legitimate and should be fixed in the official Majordomo release.
 > 
 > The fix is trivial.  Simply put a close(NEW); at a strategic place in
 > the do_unsubscribe routine.  I put it right before the block of code
 > starting with the line 'if ($match_count == 1) {'.
 > 
 > So, Brent, how about getting this bug fixed at last?
 > 
 > ---- jingoro@tcp.com -- Jim Lick -- jingoro@rahul.net -- jIngOrO@CaveMUCK ----
 > --:):-- perfect little dream the kind that hurts the most -- |\| | |/| --:(:--
 > --- CaveMUCK is back! --- Telnet to cave.tcp.com (128.95.10.106) port 2283 ---
 > -- Finger me for fun,  excitement or for my PGP Public Key Encryption Block --



References:
Indexed By Date Previous: Re: multiple unsubscribes
From: Kasuga Jingoro <jingoro@rahul.net>
Next: Re: type-cast voodoo (?)
From: Alan Millar <amillar@bolis.sf-bay.org>
Indexed By Thread Previous: Re: multiple unsubscribes
From: Kasuga Jingoro <jingoro@rahul.net>
Next: Re: multiple unsubscribes
From: Brent Chapman <brent@GreatCircle.COM>

Google
 
Search Internet Search www.greatcircle.com