[Mailman-Users] A Known bug with confirmation string & excessive bounces

[Stonewall Ballard]

On 4/8/03 11:49 PM, "Barry Warsaw" <barry at python.org> wrote:

> On Mon, 2003-04-07 at 12:18, Martin Maechler wrote:
> 
>> But I've got quite a few user reports about non-working confirmation
>> strings --- and I've experienced it myself quite a bit.  E.g., just
>> today, I tried to change one of my e-mail addresses to another one,
>> got the confirmation e-mail, but replying to it (or using the web
>> interface with the correct confirmation string) always gave errors about
>> invalid strings.
> 
> I believe we have a working theory about why this happens, even though
> we don't quite understand the root cause (perhaps a race condition in
> the LockFile.py implementation?).
> 
> Thomas Wouters implemented a fix that will show up in Mailman 2.1.2, and
> I expect to release that patch very soon now, depending on how soon I
> can finish my taxes. :)

This happens to me consistently, as I reported on Mailman-developers on
March 29. I can reproduce it at will. My lists are not very busy. A race
condition is practically impossible.

I'm using a March 29 CVS build.

Here's my message of 3/29/03:

> In 2.1.1+ (confirmed with a fresh CVS build today), when a user replies to a
> confirmation message sent by Mailman in response to an email address change
> (on the web page), Mailman changes the address, but replies with an error
> (below). This confuses users.
> 
> The subscribe log has entries for the new address as "pending", then as "new",
> with no errors or apparent problems.
> 
> It seems odd to me that this hasn't been reported as a bug, since it appears
> to have been around for several months, at least. Or is there something broken
> in my Mailman installation? I'm on OS X Server 10.2.4.
> 
> BTW, note the typo "you are not current a member".


 - Stoney