Monday, February 7, 2011

Converting from mbox to Maildir with Postfix and Dovecot on Debian

This is the procedure I used for converting from mbox to Maildir with Postfix and Dovecot running Debain 5.x. This assumes that you have individual system accounts for each mail user and that the mbox files are currently in /var/spool/mail/USER and /home/USER/mail/FOLDERS.

All the following commands/edits will need to be done as root:

Install mb2md:

aptitude install mb2md

Stop the mail servers:

/etc/init.d/postfix stop
/etc/init.d/dovecot stop


Edt the /etc/posfix/main.cf:

home_mailbox = Maildir/

Edit the /etc/procmailrc (if you are using procmail for local delivery/filtering):

ORGMAIL=$HOME/Maildir/
MAILDIR=$ORGMAIL
DEFAULT=$ORGMAIL


Edit the /etc/dovecot/dovecot.conf:

mail_location = maildir:~/Maildir

At this point I'd backup the user accounts just in case there is an issue.

Create a bash script with the following and execute it:

for i in $( ls -1 /home ); do
su -l -c 'mb2md -m' $i; ## convert /var/spool/mail/USER
su -l -c 'mb2md -s mail -R' $i;
done


At this point you will need to restart the mail servers:

/etc/init.d/dovecot start
/etc/init.d/postfix start

Friday, October 8, 2010

HTML::Template CACHE LOAD, HIT and MISS

If you are sure you have mod_perl running and are using cache_debug to diagnose issues with repeated cache LOAD and cache MISS with HTML::Template then the following information might be useful.

Digging through the HTML::Template source I discovered the following:

my @key = ($options->{filepath});
push(@key, @{$options->{path}});
push(@key, $options->{search_path_on_include} || 0);
push(@key, $options->{loop_context_vars} || 0);
push(@key, $options->{global_vars} || 0);


The key under which the cached template is stored is a md5_hex of the previous values passed to the HTML::Template->new() method.  If you are pre-loading your templates in your startup with one set of any of these parameters but are later loading the same template with another set then you will see another CACHE LOAD.

For example if you pre-cache in your startup with:

HTML::Template->new( filename => '/my/path/templates/t.html', cache => 1);

and later in a Registry cgi you create a template object with:

my $tmpl = HTML::Template->new( filename => '/my/path/templates/t.html',  cache => 1, global_vars => 1);

you will see a CACHE LOAD rather than a CACHE HIT when /my/path/templates/t.html is loaded.  The end result is that you maybe loading and caching multiple copies of the same template unnecessarily and wasting memory.

Tuesday, January 26, 2010

Correct way to install Firefox 3.6 on Ubuntu 9.10 Karmic Koala

First close all your current firefox windows. Then open a Terminal with Applications > Accessories > Terminal. At the $ prompt enter the following:

sudo add-apt-repository ppa:mozillateam/firefox-stable

When prompted for your password enter it and enter the following commands one on each line:

sudo apt-get update
sudo apt-get install firefox-3.6

Enjoy.

Friday, November 13, 2009

Can't log in to youtube with android app?

I was only able to get this to work by logging in using my google user/address and password which had been previously associated with my youtube username.

Wednesday, September 2, 2009

Finding perl packages with APT

If you are attempting to satisfy dependencies for perl software that was NOT available via your favourite repository (this is sometimes unavoidable in enterprise environments) you should AT LEAST attempt to satisfy those dependencies using said repository. The easiest solution is to install 'apt-file' from your repo and for any perl modules you can simply call:
apt-file search "LWP::UserAgent"

This will return the following list on Debian Lenny:
liblwp-useragent-determined-perl: /usr/share/man/man3/LWP::UserAgent::Determined.3pm.gz
libtest-mock-lwp-perl: /usr/share/man/man3/Test::Mock::LWP::UserAgent.3pm.gz
libwww-perl: /usr/share/man/man3/LWP::UserAgent.3pm.gz

Now you know that in order to satisfy a dependency on "LWP::UserAgent" you would need to install the 'libwww-perl' module from the repo.

If the module is not available via the repo you should next consider installing the package into a custom perl lib directory specifically for the software (and/or user account) in question. To do this read the documentation for the local::lib module in CPAN.

Ultimately your goal should be to not install non-packaged software to /usr/local and by doing so increase the maintainability of your systems.

Wednesday, August 12, 2009

perl MIME::Lite sendmail Return-Path fix/override

I'm not sure this affects the newest version of MIME::Lite; however, if you find that you are having issues with your messages being rejected or marked as spam, due to an internal host being added to the Return-Path, the following fix might work for you.

Rather than using:
my $ml = MIME::Lite-new(
    From => ...
To => ...
...
);

$ml->send();

Use the following:
$ml->send_by_sendmail(SetSender => 1);

Thursday, August 6, 2009

Find page count in TIFF or PDF with perl

The perl module you'll want to install is called PerlMagick (Image::Magick) and rather than loading the image into memory you can just use the Ping method. The Ping method returns an array with 4 elements of information for each page of a multi-page TIFF/PDF. If we divide the number of elements by 4 we know how many pages there are in the TIFF/PDF.
use Image::Magick;

my $im = Image::Magick->new();

my @ping_info = $im->Ping('fax.tif');

## If we access an @rray in a scalar context we get the element count
my $count = @ping_info / 4;

print $count;

There you have it, good luck.

UPDATE:

The example above may or may not work properly with PDF's.  Even if it does work it appears that ImageMagick makes an external call to the ghostscript (gs) executable on the system to determine the properties of the pdf.  Here is an example that will work for PDFs:
use PDF::API2;

my $pdf = PDF::API2->open('2.pdf');
print $pdf->pages();

I'm guessing that eventually ImageMagick won't be "broken" but you can use the above in the meantime.

UPDATE:

We recently updated to Debian 6.0 and GhostScript seems to have problems with PDFs rendered by various software (it takes 5 minutes for it to get a page count on a 4 page document).  My first line of defense for checking page counts in a pdf is the following for PDFs:
open(FH, "$temp_file");
for my $line (<FH>) {
    if ($line =~ m/\/Count\s+(\d+)/) {
        $page_count = $1;
    }
}
close(FH);

Don't be tempted to stop searching for the Count after the first match. Some PDFs will list a count for every page (like a page number). You'll want to use whatever the last match is for your actual count. If this code doesn't produce a page count I do the Image::Magick Ping. That appears to cover all scenarios I've come across.