Random return messages from dbmail-smtp?

Ok, I'm stumped. I'll think about it some more and see what
I can do.

On Fri, Dec 15, 2006, Larry Rosenman <lerlerctr.org> said:

> I did already, to both programs. 

> > On Behalf Of Aaron Stone:
> > I posted a test program to try to isolate the
issues, give it a shot.
> 
> > Aaron
_______________________________________________
DBmail mailing list
DBmaildbmail.org
htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail

Sorry if I'm stating the obvious here, but the problem is
that PG 
doesn't accept "1166132546-05-07 15:42:45" as a
valid timestamp because 
it isn't.  The question is, where did that malformed year
come from? 
This sounds like DBMail (or gmime) isn't interpreting the
date value 
correctly from the message headers.  From a quick look at
the bad 
message posted to mantis, I don't see any dates that look
particularly 
bad.  Is there a way to have gmime open that message and see
what string 
it failing with?

Matt


Aaron Stone wrote:
> Ok, I'm stumped. I'll think about it some more and see
what I can do.
> 
> On Fri, Dec 15, 2006, Larry Rosenman <lerlerctr.org> said:
> 
>> I did already, to both programs. 
> 
>>> On Behalf Of Aaron Stone:
>>> I posted a test program to try to isolate the
issues, give it a shot.
>>> Aaron
> _______________________________________________
> DBmail mailing list
> DBmaildbmail.org
> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
> 
_______________________________________________
DBmail mailing list
DBmaildbmail.org
htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail

The query we see in Larry's log files is only present in the

db_insert_physmessage_with_internal_date function. The two
test programs I
wrote trace back from char2date_str which is present in that
function.
Since it didn't fail, I'm at a loss right now to understand
what's broken.

Aaron

On Fri, Dec 15, 2006, Matthew O'Connor <matthewzeut.net> said:

> Sorry if I'm stating the obvious here, but the problem
is that PG 
> doesn't accept "1166132546-05-07 15:42:45" as
a valid timestamp because 
> it isn't.  The question is, where did that malformed
year come from? 
> This sounds like DBMail (or gmime) isn't interpreting
the date value 
> correctly from the message headers.  From a quick look
at the bad 
> message posted to mantis, I don't see any dates that
look particularly 
> bad.  Is there a way to have gmime open that message
and see what string 
> it failing with?
> 
> Matt
> 
> 
> Aaron Stone wrote:
>> Ok, I'm stumped. I'll think about it some more and
see what I can do.
>> 
>> On Fri, Dec 15, 2006, Larry Rosenman <lerlerctr.org> said:
>> 
>>> I did already, to both programs. 
>> 
>>>> On Behalf Of Aaron Stone:
>>>> I posted a test program to try to isolate
the issues, give it a shot.
>>>> Aaron
>> _______________________________________________
>> DBmail mailing list
>> DBmaildbmail.org
>> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
>> 
> _______________________________________________
> DBmail mailing list
> DBmaildbmail.org
> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
> 

-- 



_______________________________________________
DBmail mailing list
DBmaildbmail.org
htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail

Right, but the test program (as best I can tell) calls
char2date_str 
with a date string created by a call to the OS which will
most likely 
give us a sane representation.  I think the problem is that
gmime (or 
what ever code it is in DBMail that interprets the headers)
is getting 
confused by the actual string in his email.

Apologies if my comments are off the mark since I'm not
familiar with 
the code that generates the insert statement.


Matt

Aaron Stone wrote:
> The query we see in Larry's log files is only present
in the 
> db_insert_physmessage_with_internal_date function. The
two test programs I
> wrote trace back from char2date_str which is present in
that function.
> Since it didn't fail, I'm at a loss right now to
understand what's broken.
> 
> Aaron
> 
> On Fri, Dec 15, 2006, Matthew O'Connor <matthewzeut.net> said:
> 
>> Sorry if I'm stating the obvious here, but the
problem is that PG 
>> doesn't accept "1166132546-05-07
15:42:45" as a valid timestamp because 
>> it isn't.  The question is, where did that
malformed year come from? 
>> This sounds like DBMail (or gmime) isn't
interpreting the date value 
>> correctly from the message headers.  From a quick
look at the bad 
>> message posted to mantis, I don't see any dates
that look particularly 
>> bad.  Is there a way to have gmime open that
message and see what string 
>> it failing with?
>>
>> Matt
>>
>>
>> Aaron Stone wrote:
>>> Ok, I'm stumped. I'll think about it some more
and see what I can do.
>>>
>>> On Fri, Dec 15, 2006, Larry Rosenman
<lerlerctr.org> said:
>>>
>>>> I did already, to both programs. 
>>>>> On Behalf Of Aaron Stone:
>>>>> I posted a test program to try to
isolate the issues, give it a shot.
>>>>> Aaron
>>> _______________________________________________
>>> DBmail mailing list
>>> DBmaildbmail.org
>>> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
>>>
>> _______________________________________________
>> DBmail mailing list
>> DBmaildbmail.org
>> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
>>
> 
_______________________________________________
DBmail mailing list
DBmaildbmail.org
htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail

On Fri, 15 Dec 2006, Matthew O'Connor wrote:

> Right, but the test program (as best I can tell) calls
char2date_str with a 
> date string created by a call to the OS which will most
likely give us a sane 
> representation.  I think the problem is that gmime (or
what ever code it is 
> in DBMail that interprets the headers) is getting
confused by the actual 
> string in his email.

Right.  I saw this failure a LOT yesterday.  The "Bad
Message" was one that 
I KNOW failed from the timestamps in the maillog and
postgres log.

I'm testing dbmail by having my personal Exim filter pipe
the message
to dbmail-smtp as an unseen.  This message was the one that
was written to
the MBX format file that I've always written to.  I'm not
sure of any
sanity/other changes that are made on the fly by Exim as it
goes to
deliver the message via it's MBX driver, vs the pipe driver.

I would see these failures randomly on messages. 
Unfortunately, I don't
know of a way to capture what dbmail-smtp is seeing when it
fails.  :(

I can NOT afford to have dbmail-smtp generate the bounces
like it was doing
yesterday, as I'll be unsubscribed from the list(s) I'm on.

What *CAN* I do to help here?

LER

>
> Apologies if my comments are off the mark since I'm not
familiar with the 
> code that generates the insert statement.
>
>
> Matt
>
> Aaron Stone wrote:
>> The query we see in Larry's log files is only
present in the 
>> db_insert_physmessage_with_internal_date function.
The two test programs I
>> wrote trace back from char2date_str which is
present in that function.
>> Since it didn't fail, I'm at a loss right now to
understand what's broken.
>> 
>> Aaron
>> 
>> On Fri, Dec 15, 2006, Matthew O'Connor
<matthewzeut.net> said:
>> 
>>> Sorry if I'm stating the obvious here, but the
problem is that PG doesn't 
>>> accept "1166132546-05-07 15:42:45" as
a valid timestamp because it isn't. 
>>> The question is, where did that malformed year
come from? This sounds like 
>>> DBMail (or gmime) isn't interpreting the date
value correctly from the 
>>> message headers.  From a quick look at the bad
message posted to mantis, I 
>>> don't see any dates that look particularly bad.
 Is there a way to have 
>>> gmime open that message and see what string it
failing with?
>>> 
>>> Matt
>>> 
>>> 
>>> Aaron Stone wrote:
>>>> Ok, I'm stumped. I'll think about it some
more and see what I can do.
>>>> 
>>>> On Fri, Dec 15, 2006, Larry Rosenman
<lerlerctr.org> said:
>>>> 
>>>>> I did already, to both programs. 
>>>>>> On Behalf Of Aaron Stone:
>>>>>> I posted a test program to try to
isolate the issues, give it a shot.
>>>>>> Aaron
>>>>
_______________________________________________
>>>> DBmail mailing list
>>>> DBmaildbmail.org
>>>> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
>>>> 
>>> _______________________________________________
>>> DBmail mailing list
>>> DBmaildbmail.org
>>> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
>>> 
>> 
> _______________________________________________
> DBmail mailing list
> DBmaildbmail.org
> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
>

-- 
Larry Rosenman                     http://www.lerctr.org/~ler

Phone: +1 512-248-2683                 E-Mail: lerlerctr.org
US Mail: 430 Valona Loop, Round Rock, TX 78681-3893
_______________________________________________
DBmail mailing list
DBmaildbmail.org
htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail

No no, you're right. I do have more functions to look at.

On Fri, Dec 15, 2006, Matthew O'Connor <matthewzeut.net> said:

> Right, but the test program (as best I can tell) calls
char2date_str 
> with a date string created by a call to the OS which
will most likely 
> give us a sane representation.  I think the problem is
that gmime (or 
> what ever code it is in DBMail that interprets the
headers) is getting 
> confused by the actual string in his email.
> 
> Apologies if my comments are off the mark since I'm not
familiar with 
> the code that generates the insert statement.
> 
> 
> Matt
> 
> Aaron Stone wrote:
>> The query we see in Larry's log files is only
present in the 
>> db_insert_physmessage_with_internal_date function.
The two test programs I
>> wrote trace back from char2date_str which is
present in that function.
>> Since it didn't fail, I'm at a loss right now to
understand what's broken.
>> 
>> Aaron
>> 
>> On Fri, Dec 15, 2006, Matthew O'Connor
<matthewzeut.net> said:
>> 
>>> Sorry if I'm stating the obvious here, but the
problem is that PG 
>>> doesn't accept "1166132546-05-07
15:42:45" as a valid timestamp because 
>>> it isn't.  The question is, where did that
malformed year come from? 
>>> This sounds like DBMail (or gmime) isn't
interpreting the date value 
>>> correctly from the message headers.  From a
quick look at the bad 
>>> message posted to mantis, I don't see any dates
that look particularly 
>>> bad.  Is there a way to have gmime open that
message and see what string 
>>> it failing with?
>>>
>>> Matt
>>>
>>>
>>> Aaron Stone wrote:
>>>> Ok, I'm stumped. I'll think about it some
more and see what I can do.
>>>>
>>>> On Fri, Dec 15, 2006, Larry Rosenman
<lerlerctr.org> said:
>>>>
>>>>> I did already, to both programs. 
>>>>>> On Behalf Of Aaron Stone:
>>>>>> I posted a test program to try to
isolate the issues, give it a shot.
>>>>>> Aaron
>>>>
_______________________________________________
>>>> DBmail mailing list
>>>> DBmaildbmail.org
>>>> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
>>>>
>>> _______________________________________________
>>> DBmail mailing list
>>> DBmaildbmail.org
>>> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
>>>
>> 
> _______________________________________________
> DBmail mailing list
> DBmaildbmail.org
> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
> 

-- 



_______________________________________________
DBmail mailing list
DBmaildbmail.org
htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail

Larry Rosenman wrote:
> On Fri, 15 Dec 2006, Matthew O'Connor wrote:
> 
>> Right, but the test program (as best I can tell)
calls char2date_str 
>> with a date string created by a call to the OS
which will most likely 
>> give us a sane representation.  I think the problem
is that gmime (or 
>> what ever code it is in DBMail that interprets the
headers) is getting 
>> confused by the actual string in his email.
> 
> Right.  I saw this failure a LOT yesterday.  The
"Bad Message" was one 
> that I KNOW failed from the timestamps in the maillog
and postgres log.
> 
> I'm testing dbmail by having my personal Exim filter
pipe the message
> to dbmail-smtp as an unseen.  This message was the one
that was written to
> the MBX format file that I've always written to.  I'm
not sure of any
> sanity/other changes that are made on the fly by Exim
as it goes to
> deliver the message via it's MBX driver, vs the pipe
driver.
> 
> I would see these failures randomly on messages. 
Unfortunately, I don't
> know of a way to capture what dbmail-smtp is seeing
when it fails.  :(
> 
> I can NOT afford to have dbmail-smtp generate the
bounces like it was doing
> yesterday, as I'll be unsubscribed from the list(s) I'm
on.
> 
> What *CAN* I do to help here?

<disclaimer> I haven't played with DBMail code much at
all so use 
appropriate grain of salt </disclaimer>

Aaron said that insert statment is only created from on
place:
db_insert_physmessage_with_internal_date
so I would start there, either use gdb if you can or more
low-tech, 
start putting in some print statements to see what is going
on.  If you 
can narrow down the problem, I'm sure Aaron or Paul will be
able to nail it.

Matt
_______________________________________________
DBmail mailing list
DBmaildbmail.org
htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail

On Fri, 15 Dec 2006, Matthew O'Connor wrote:

> Larry Rosenman wrote:
>> On Fri, 15 Dec 2006, Matthew O'Connor wrote:
>> 
>>> Right, but the test program (as best I can
tell) calls char2date_str with 
>>> a date string created by a call to the OS which
will most likely give us a 
>>> sane representation.  I think the problem is
that gmime (or what ever code 
>>> it is in DBMail that interprets the headers) is
getting confused by the 
>>> actual string in his email.
>> 
>> Right.  I saw this failure a LOT yesterday.  The
"Bad Message" was one that 
>> I KNOW failed from the timestamps in the maillog
and postgres log.
>> 
>> I'm testing dbmail by having my personal Exim
filter pipe the message
>> to dbmail-smtp as an unseen.  This message was the
one that was written to
>> the MBX format file that I've always written to. 
I'm not sure of any
>> sanity/other changes that are made on the fly by
Exim as it goes to
>> deliver the message via it's MBX driver, vs the
pipe driver.
>> 
>> I would see these failures randomly on messages. 
Unfortunately, I don't
>> know of a way to capture what dbmail-smtp is seeing
when it fails.  :(
>> 
>> I can NOT afford to have dbmail-smtp generate the
bounces like it was doing
>> yesterday, as I'll be unsubscribed from the list(s)
I'm on.
>> 
>> What *CAN* I do to help here?
>
> <disclaimer> I haven't played with DBMail code
much at all so use appropriate 
> grain of salt </disclaimer>
>
> Aaron said that insert statment is only created from on
place:
> db_insert_physmessage_with_internal_date
> so I would start there, either use gdb if you can or
more low-tech, start 
> putting in some print statements to see what is going
on.  If you can narrow 
> down the problem, I'm sure Aaron or Paul will be able
to nail it.
The problem is some fail, some work, so I have no way of
knowing.

HOWEVER, now that I have the "bad" message, I will
try and play with it under
the debugger tonight/this weekend.

LER
>
> Matt
> _______________________________________________
> DBmail mailing list
> DBmaildbmail.org
> htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail
>

-- 
Larry Rosenman                     http://www.lerctr.org/~ler

Phone: +1 512-248-2683                 E-Mail: lerlerctr.org
US Mail: 430 Valona Loop, Round Rock, TX 78681-3893
_______________________________________________
DBmail mailing list
DBmaildbmail.org
htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail

Ya know, this is a royal pain.  I tried running the message
that failed with
gdb, and it worked.

There is something weird going on here.....

I'm going to see what I can do to trap this error....

Suggestions on what/where to put some debugging would be
appreciated!

LER


-- 
Larry Rosenman                     http://www.lerctr.org/~ler

Phone: +1 512-248-2683             E-Mail: lerlerctr.org
US Mail: 430 Valona Loop, Round Rock, TX 78681-3893


_______________________________________________
DBmail mailing list
DBmaildbmail.org
htt
ps://mailman.fastxs.nl/mailman/listinfo/dbmail