...strangely, the message makes it through and is noted by the bug tracker.

Oh, and David thinks it's happening to everybody.  Any thoughts, Mark?

Eric,

I'm really not sure. We could try tracing the flow of e-mail, and also reviewing
the ChangeLog of newer Round-up releases to see if there's something that looks
like a related bug fix. 

I'm not much of an expert with Round-Up, Python or the e-mail system in use,
unfortunately. 

  Mark

This message is an attempt to debug the bouncing trouble.
-- 
David Roundy
Department of Physics
Oregon State University

Tonight I reviewed the mail flow for bugs@darcs.net. Here's some details from that: 

- The mail is managed by exim4. 
- It appears that the bugs address is managed throug here:
  /etc/exim4/conf.d/router/459_bugs

Which has these contents:
darcs_bugs:
  driver = redirect
  local_parts = darcs-bugs
  senders = bugs@darcs.net : bugs-bounces@darcs.net
  user = www-data
  data = "|/usr/bin/procmail /etc/darcs_bugs_procmailrc"
  pipe_transport = address_pipe
#####

The procmailrc file it refers to looks like this:
###
:0 chfw
SUBJECT=| formail -xSubject: | perl -pe 's/\[darcs-[^\]]+\]//'

:0 hfw
| /usr/bin/formail -i "Subject: $SUBJECT"

:0
|/usr/bin/roundup-mailgw /var/lib/roundup/trackers/darcs

####

I found it notable that Roundup includes documentation for integrating with Exim
( search for Exim on this page:   
  http://roundup.sourceforge.net/doc-1.0/installation.html ).

However, the examples they cite there have noticeable differences from the
system we are using. 

So, without fully understanding how Roundup or Exim work, I advise evaluating
configuration changes that would bring us more inline with the official
Roundup/Exim example. 

I hope that might solve the dupe bug reports that get happen frequently now.

On Thu, Feb 14, 2008 at 04:44:06AM -0000, Mark Stosberg wrote:
> However, the examples they cite there have noticeable differences from the
> system we are using. 
> 
> So, without fully understanding how Roundup or Exim work, I advise evaluating
> configuration changes that would bring us more inline with the official
> Roundup/Exim example. 

We've got a major difference, which is that we filter all our bugs email
through a mailman mailing list in order to achieve moderation with a
reasonable user-interface for the moderator (me) that includes
whitelisting.  Perhaps this is no longer necessary, as it seems that the
bugs@darcs.net only gets one spam/day or less (I've no idea why,
darcs-devel and darcs-users both get far more spam).

It also might work this way because I've been hesitant to change a working
configuration.  You're welcome to make (cautious) changes to the exim
setup, or to make more specific suggestions.

> I hope that might solve the dupe bug reports that get happen frequently now.

That would be nice.
-- 
David Roundy
Department of Physics
Oregon State University

My personal hope is that with my initial research, someone else will see the bug. :)

I'll study it some more, but will be hesitant to make changes until I under
understand what's going. If I do change anything, I'll have good backups and
documentation. 

   Mark

I can look at this a little more.

If this provides any sort of clue, darcs-devel seems to get second bug report,
not the first one.

Eric, that clue may have given me the answer:  it looks like bugs is subscribed
to darcs-devel and may have been getting an extra copy sent by darcs-devel. 
I've left bugs subscribed to darcs-devel (so that bug email messages will be
delivered) but set it to not receive any darcs-devel traffic.  Maybe this will
fix it?

David

Well, here is my reply via email...

2008/6/5 David Roundy <bugs@darcs.net>:
> Eric, that clue may have given me the answer:  it looks like bugs is subscribed
> to darcs-devel and may have been getting an extra copy sent by darcs-devel.
> I've left bugs subscribed to darcs-devel (so that bug email messages will be
> delivered) but set it to not receive any darcs-devel traffic.  Maybe this will
> fix it?

Hmm... still get a failed tracker submission if I reply via email.
And as per issue908 and issue909, we're still getting duplicates.
Oh well...

Whoops! Benjamin just provided a counter-example to my
darcs-devel-gets-second-bug observation (issue918 and issue919).  Hmm, I wonder
if this is related to bugs@darcs.net no longer receiving its messages, or if my
observation was just wrong.

Another clue is that the bugtracker seems to have two entries for every user. 
Look up marlowsd, for example

Can I take a look at the exim/mailman/roundup setup ? I have experience with
debugging twisty email & python setups and might be able to help.

For the record: I did get this access and am looking at the setup on  
and off. This reply will test mail-in and probably show me a bounce.

There are a few problems discussed on this page. Focussing on the "bounce when
mailing in issue", it seems to sometimes just bounce without updating the
tracker. Also I seem to be blocked on this until I can get a bit more access to
exim logs and to edit roundup config and code, to increase logging. Until then I
can't trace the flow, but I will study the code a bit more in case something
jumps out.

Mark! Help! :-)

Can you give Simon the needed access to the exim logs and roundup config (or
maybe dig up whatever information he needs?) 

[now new submissions seem to get 2 bounces and no entries on the tracker]

David took care of this, so I can see more. I'll return to this asap.

test 1

test 2, expecting no bounce.

test 3, let's see

Jason: much better. :) Thanks. With luck this will fix the duplicate  
new issues as well.

Duplicated new issues is also fixed. Re other thing mentioned on this page,
duplicate users, let's open a new issue if needed. I see that some users have
duplicates, but it doesn't seem to be hurting us yet. Closing this issue.

On Thu, Aug 14, 2008 at 20:52:20 -0000, Simon Michael wrote:
> Duplicated new issues is also fixed. Re other thing mentioned on this page,
> duplicate users, let's open a new issue if needed. I see that some users have
> duplicates, but it doesn't seem to be hurting us yet. Closing this issue.

Just one more (minor) thing Simon...

If we're adding darcs-devel to the nosy list, I think this means we need
to stop having a special rule to forward new entries to darcs-devel as
well.

Not nearly as serious as the issue you fixed, however.

Marked resolved again.

Also, as per issue999, I am censoring the messages below and removing them from
the bugtracker

=== msg 5517 censored ===
On Thu, Aug 14, 2008 at 12:43 PM, Simon Michael <bugs[REDACTED]> wrote:

>
> Simon Michael <simon[REDACTED]> added the comment:
>
> I have attached the exim and roundup logs from handling a reply to this
> issue.
> You can see that roundup is receiving the mail twice, the first is
> processed
> normally and the second fails "due to concurrent update". If I am reading
> the
> exim log rightly this machine is subscribed to darcs devel as
> darcs-bugs[REDACTED] and as darcs-bugs[REDACTED]. I think we need
> a
> list admin to remove the darcs-bugs[REDACTED] subscriber from the
> list(s).

I found darcs-bugs@ subscribed to the bugs list and removed it from there.
That's the only place I could find it though.

=== msg5514 censored ===

I have attached the exim and roundup logs from handling a reply to this issue.
You can see that roundup is receiving the mail twice, the first is processed
normally and the second fails "due to concurrent update". If I am reading the
exim log rightly this machine is subscribed to darcs devel as
darcs-bugs[REDACTED (darcs)] and as darcs-bugs[REDACTED (osu)]. I think we need a
list admin to remove the darcs-bugs[REDACTED (osu)] subscriber from the list(s).