Ask Bjørn Hansen | 1 Oct 2007 08:28
Gravatar

Re: more reset_connection() and uninitialized values


On Sep 30, 2007, at 10:51 AM, Hanno Hecker wrote:

[prefork reset problem]
> ...and on the second connection:
> $ telnet hex 2525
> Trying 10.7.7.2...
> Connected to hex.city.ankh-morp.org.
> Escape character is '^]'.
> 220 hex ESMTP qpsmtpd 0.42rc1 ready; send us your mail, but not  
> your spam.
> ehlo hex
> 503 but you already said HELO ...
>
> Mhh, seems like it does it always like this, no matter which order...

Is this a regression from 0.40?   If it is then we need to fix it  
before releasing 0.42 - if it happened in 0.40 too then we can fix it  
in 0.43.   (I guess prefork was broken enough in 0.40 that it doesn't  
really matter either way?).

  - ask

--

-- 
http://develooper.com/ - http://askask.com/

Anilkumar | 1 Oct 2007 12:30
Picon
Favicon

Reply ID & from Name

Dear All,
   I am intrested in knowing the From Name of mail
sender & his reply e-mail id,

what will be the command for it.

Anil

      Unlimited freedom, unlimited storage. Get it now, on http://help.yahoo.com/l/in/yahoo/mail/yahoomail/tools/tools-08.html/

Hanno Hecker | 1 Oct 2007 20:56
Picon

Re: more reset_connection() and uninitialized values

On Sun, 30 Sep 2007 23:28:42 -0700
Ask Bjørn Hansen <ask <at> develooper.com> wrote:
> > Mhh, seems like it does it always like this, no matter which order...
> 
> Is this a regression from 0.40?   If it is then we need to fix it  
> before releasing 0.42 - if it happened in 0.40 too then we can fix it  
> in 0.43.   (I guess prefork was broken enough in 0.40 that it doesn't  
> really matter either way?).
This does not happen in 0.40. I think it belongs to the transaction /
connection id stuff. We should fix it before 0.42, the known bugs from
0.40 in prefork are fixed.

	Hanno

Johan Almqvist | 3 Oct 2007 09:07
Favicon
Gravatar

Re: Reply ID & from Name

Anilkumar wrote:
> Dear All,
>    I am intrested in knowing the From Name of mail
> sender & his reply e-mail id,
> 
> what will be the command for it.

$transaction->header->get("From")

-Johan
--

-- 
Johan Almqvist
Christies gate 34 A
0557 Oslo
Norway

Mobile: +47 40 04 68 21
E-mail: johan <at> almqvist.net

sub-194-4 | 3 Oct 2007 18:26
Favicon

Too many open files at lib/Qpsmtpd.pm

qmsptmd-forkserver 0.32

I am seeing this error, increasingly, but far from being for every
connection:

could not open configfile ./config/logging: Too many open files at
lib/Qpsmtpd.pm line 168, <GEN3262> line 1430.

The last item is sometimes different. I presume the last item is another
process qpsmtpd is trying to run. When this happens for a connection, I
may end up with hundreds of these lines in the log file.

Is there some place I should be increasing the number of open files
allowed? Where and how? Thanks.

Chris Lewis | 3 Oct 2007 20:09
Favicon

Hi, bug report/feature request

Running 0.40 qpsmtpd-async as a spamtrap on both Linux and Solaris 
(2.8/2.9).

[I use ordinary poll on Linux because my OS is too old]

I've hacked qpsmtpd-async to support daemonization and pid locking 
files.  Rather than dig into SVN, who should I send the modified copy 
to?  Matt?

I note that plugins/tls_cert clobbers the machine's name on Solaris.  It 
issues `hostname --fqdn` in an attempt to find the machine's full name, 
and on Solaris, that's treated as setting the hostname to "--fqdn". 
Ouch.  Unfortunately, this OS is obnoxious, and tends ot have 
non-qualified hostnames (under hostname, and Sys::Hostname).  So, not 
sure what to suggest here.

John Peacock | 3 Oct 2007 20:28
Picon

Re: Hi, bug report/feature request

Chris Lewis wrote:
> I note that plugins/tls_cert clobbers the machine's name on Solaris.  It 
> issues `hostname --fqdn` in an attempt to find the machine's full name, 
> and on Solaris, that's treated as setting the hostname to "--fqdn". 
> Ouch.  

Sorry, that's my fault (but in my defense, you shouldn't be running that 
script as root in any case).  I'm sure that there is a 
platform-independent way to get the hostname, but for the moment you can 
just call the script (not as root) with

	./plugins/tls_cert --CN your.host.name

and the cert will still be created (or you can pass in all of the other 
options as well, if you care what your cert looks like).  I'll look to 
protect that `hostname` call...

John

Robin Bowes | 3 Oct 2007 21:00
Favicon

Re: Hi, bug report/feature request

John Peacock wrote:
> 
> I'm sure that there is a platform-independent way to get the hostname

How about:

use Sys::Hostname;
$host = hostname;

R.

Chris Lewis | 3 Oct 2007 21:06
Favicon

Re: Hi, bug report/feature request

Robin Bowes wrote:
> John Peacock wrote:
>> I'm sure that there is a platform-independent way to get the hostname
> 
> How about:
> 
> use Sys::Hostname;
> $host = hostname;

Returns the non-qualified hostname (at least on this Solaris).  I guess 
when it boils down to it, this machine is misconfigured.  Sigh.

Chris Lewis | 3 Oct 2007 21:10
Favicon

Re: Hi, bug report/feature request

John Peacock wrote:
> Chris Lewis wrote:
>> I note that plugins/tls_cert clobbers the machine's name on Solaris.  
>> It issues `hostname --fqdn` in an attempt to find the machine's full 
>> name, and on Solaris, that's treated as setting the hostname to 
>> "--fqdn". Ouch.  
> 
> Sorry, that's my fault (but in my defense, you shouldn't be running that 
> script as root in any case).

True enough.

> I'm sure that there is a 
> platform-independent way to get the hostname, but for the moment you can 
> just call the script (not as root) with
> 
>     ./plugins/tls_cert --CN your.host.name

[Aside from the error hostname will throw.]
> 
> and the cert will still be created (or you can pass in all of the other 
> options as well, if you care what your cert looks like).  I'll look to 
> protect that `hostname` call...

Any idea why I'd be getting these:

_convert_to_ssl failed (Could not create SSL socket: Socket operation on 
non-socket at ./plugins/tls line 158

[At end of _convert_to_ssl(), I'm logging $ <at>  if it's non-null]
(Continue reading)


Gmane