LinuxQuestions.org - Blogs - dr_agon

Which MUA for Linux?

dr_agon — Sun, 09 Dec 2012 13:40:13 GMT

applies to: linux, KMail, Opera, Balsa, Claws

Recently I got sick of new KMail. I tried to be patient and after upgrading to new KDE4 (together with Kubuntu 12.04, which is "Long Term Support", so it was ment to be "stable") I ran trough all pains with migrating my e-mails and configuring it together with Akonadi, Nepomuk, and Virtuoso. I hoped to get the whole KDE-PIM usable again. Now, after half a year, I gave up and searched for replacement.
I needed Mail User Agent with graphical interface (GUI), POP3 support, ability to store locally several thousands of e-mails, have not only spam but other filter capabilities, and support custom message tags to help me organize messsages.
Here are options I considered with short comments about each program:

1. KMail2 (the last version I used was 4.9.4)
It theoretically has all required features and works well with other parts of KDE-PIM. Unfortunately, it has major bugs and disadvantages:

It caused some e-mail loss for me, in unknown and not repeatable situations, so I could not file a bug for it.
It installs and uses separate MySQL database for storing metadata about e-mails, which consumes resources (well, it is really Akonadi, but you cannot use KMail without it). It may be possible to combine it with system-wide MySQL installation, but due to unstable and unpredictable behavior of KMail - I did't try it.
For simple task of custom tagging of messages it requires Nepomuk Semantic Desktop running - another huge, resource-greedy subsystem. Couple of times all my tags just disappeared, leaving my hundreds of mails bare.
Due to asynchronous processing it frequently asks user for resolving conflicts due to trivial problems. For example when spam filter tries to move message to another folder, and I managed to view the message during "filtering", one version of message is "unread, spam", but other becomes "read" and the poor KMail can't manage it itself. I'd rather wait until it finishes filtering, but it reports finishing right after queuing the messages for filter, so how can I know when the filtering ends?
The annoying "synchronizing folder" messages seem just a small inconvenience in this circumstances.

2. Opera Mail
Opera is my primary browser, so I decided to give OperaMail a try. I knew about its convenient message tagging(=labelling) method, and I think this is the main advantage of OperaMail. But it lacks other important things:

It does not support filtering message through external program.
It seems that Opera uses its own spam detection mechanism, which would require training from the beginning, and probably would not be transferable.
The thing that convinced me to drop OperaMail was the format of message folders. It is neither mbox, maildir, nor mh. I asked myself - what if I wanted to switch from Opera to other mail client - how would I import messages FROM OperaMail to another MUA? Manually copying?

3. Balsa (http://pawsa.fedorapeople.org/balsa/features.html)
Its feature list looked promising.

Since it supports maildir I was able to configure it to use the same directories as KMail (local mail folder), so the problem of importing all my mails just disappeared. This was a nice surprise. In fact I could use both KMail and Balsa with the same maildir. And Balsa was soooo much faster!
The filtering capabilities were sufficient, but to get best results one would probably need procmail.
It was easy configurable, and had many useful options.
Yet, Balsa doesn't support custom message tags. The message can be either "flagged" or "unflagged", no more. Maybe when this gets implemented, I'll try it again.

4. Claws mail (http://www.claws-mail.org/)

I installed version 3.9.0 from Ubuntu PPA. This was convenient.
Then, I had to import messages from KMail to MH folder. This was done easily using a script.
There are lots of filtering and processing options which may seem confusing at the beginning, but overall it can be very useful.
I managed to configure filters to use qsf as spam recognition program, so my well trained database remained valid!
Creating and adding custom tags to messages was flawless, and that's what I rely on in my work.
Searching through big folders is quick.

The overall performance of Claws Mail is very pleasant in comparison to KMail, and for now it remains my e-mail client.

Giving up KMail, I also searched for replacement of KCalendar. This is another story, but just to mention it shortly, I use Chandler.

Where are desktop icons stored in KDE4 Plasma?

dr_agon — Sun, 06 Mar 2011 12:34:44 GMT

After switching to KDE4 I was little confused with its new philosophy of Plasma containers and activities. As it looked interesting, I tried to find more information. Now, there are some nice posts showing how to use it, but I couldn't find the answer for a simple question: where the information about desktop icons is stored on a disk?

Just for the record, in KDE 4.4.5 it is in
'~/.kde/share/config/plasma-desktop-appletsrc'

batch resize images using find, xargs and convert

dr_agon — Sun, 20 Feb 2011 22:30:43 GMT

applies to: linux, find, xargs, convert, image processing

'convert' command, which is a part of ImageMagick package can be used for image resizing, but it lacks ability to process multiple files (in version of ImageMagick 6.5.7-8 2010-12-02). Instead of writing a shell script you can use find and xargs commands to collect filenames and invoke convert with one filename at a time, like this:

Code:

find -iname "*.jpg" -printf "%f\0" \
 | xargs -0 -I {} convert -resize 50% {} small/{}

The general form would be:

Code:

find [directory to search, if omitted = current] -iname [search pattern] -printf "%f\0" \
 | xargs -0 -I {} convert -resize [% or widthxheight] [directory for converted images]/{}

Little explanation:
The find parameters make search case-insensitive and output NULL delimited filenames without path. The xargs -I {} parameter makes {} a "placeholder" for data from find.
See man pages for details.

Differencies between OpenOffice.org under Windows and Linux

dr_agon — Tue, 02 Dec 2008 21:53:44 GMT

applies to OpenOffice.org, Linux, Windows, Python, XML

Recently I try to develop modules for already existing, big project using Python and OpenOffice.org. The application is meant to work both under Linux and Windows. Unfortunately, it is not easy.

Python-UNO bridge to OpenOffice, which is part of OO suite, is different. Under Linux it works flawlessly with the fairly new Python v. 2.5. Under Windows it works only with supplied Python v. 2.3.4. It wouldn't be so bad if recent wxWidgets worked with this version, but they don't.
Under Linux OpenOffice.org is shipped with 'OpenDocument Text Flat XML' filter for importing/exporting .fodt files. I don't know why, but Windows version lacks this filter. I did not find this filter available for download anywhere. The solution is simple, but not obvious: you can export this filter from Linux version and import under Windows. (Many thanks to r4zoli for this suggestion on the forum.

Now, I cannot manipulate OO documents using UNO, but I can prepare XML files which are properly loaded into OO both under Linux and Windows, and are very easy to generate and change. I must say I like XML.

Avoiding wxPython PrintPreview segmentation fault or double free error

dr_agon — Sat, 29 Nov 2008 02:23:40 GMT

applies to: Python, wxPython (v.2.8), wxWidgets, wx.PrintPreview, wx.Printout

When creating wx.PrintPreview class, you must pass 2 parameters of wxPrintout class (see doc)
If someone (like me) gets a silly idea of passing the same instance of wxPrintout class as both parameters, on closing the preview frame he will get segmentation fault or double free error.
Example (don't do this!):

Code:

class MyPrintout(wx.Printout):
  
  #rest of code
  
printout = MyPrintout(MyParameters) #  instance of class based on wx.Printout 
printpreview = wx.PrintPreview(printout, printout, MyPrintData)

Even if both parameters must refer to the same class, they should be passed as different instances, like this:

Code:

class MyPrintout(wx.Printout):
  
  #rest of code
  
printout_preview = MyPrintout(MyParameters) # first instance of class based on wx.Printout 
printout_printing = MyPrintout(MyParameters) # second instance of class based on wx.Printout 
printpreview = wx.PrintPreview(printout_preview, printout_printing, MyPrintData)

Now, it seems logical to me, but cost me lots of time to figure it out.

Setting udev permission for raw host disk access by virtualbox machine.

dr_agon — Tue, 14 Oct 2008 14:02:16 GMT

applies to: linux, udev, permissions, device node, virtualbox

To allow raw host disk access by guest machine on VirtualBox the user must have read-write permission to all partitions which are to be accessed. You can do this by something like

Code:

sudo chmod a+rw /dev/sda7

but executing it every time you need to boot virtual machine is annoying. You may want to set the permissions at boot time.
In modern kernels (since 2.6) you can use the udev rules for setting default permissions to selected devices. I created a file /etc/udev/rules.d/60-whatever.rules with a following rule:

Code:

KERNEL=="sda[78]*",		MODE="0666"

This sets the permission for everyone to read-write /dev/sda7 and /dev/sda8.

That's all I needed.

How to remove selected messages left on the POP3 server.

dr_agon — Mon, 13 Oct 2008 21:46:01 GMT

applies to: KMail, Evolution, Perl, POP3, mail client

Unfortunatelly my favourite mail client (Kmail) cannot leave messages on the POP3 server while retrieving them and delete them later, eg. when they are moved to 'Trash" folder (like Thunderbird). The same applies to Evolution. Of course I know there are IMAP servers, but that's not the solution I like most. I also did not want to periodically erase all messages left on POP3 server. I did not find any reasonable solution, so I made it myself. It is available as DeleteFromPOP3 project at Sourceforge.net(tm).

The idea is simple:

mark messages for deleting in the mail client - It is not so difficult to do it manually. I just delete mails from my inbox or mark them as spam. Then, from time to time I mark all files in the trash and spam folder for removal from POP3 server. When you create a mail filter and add it to the toolbar, it just requires one click to mark messages for removal.
pass the list of messages marked for deletion to the external program - This is done together with marking for removal. No additional action required.
run external program to delete messages from the server - You can run it automatically if you wish.

To achieve this, I wrote two Perl scripts, and created a mail filter in Kmail.
How it works

A filter is invoked which executes a helper script passing the message as input.
A message from mail client is parsed and information needed for
identification of message on the server is collected.
The script produces a pair of and for
every message.
The pairs of and are stored in
separate file, each in separate line
The second script, which actually deletes messages from POP3
server, is executed whenever you like.
This script requires a list of messages created by previous
script.
It also requires POP3 account details to connect to server(s).
After connecting to the POP3 server it looks through all
messages and checks if a message is on the list and should be
deleted.
After deleting messages it truncates the message list file.

This may not be the best solution, but it works for me, and there is a good chance it will work for others, too. Any comments are welcomed.

Enjoy.

CVS diff default behavior

dr_agon — Mon, 13 Oct 2008 20:55:48 GMT

applies to: linux, cvs, diff

I spent some time wondering why cvs diff does not show the local files which differ from the most recent revision. It turned out that this is simply the default behavior, as described in man page:

Quote:

The default action is to compare your working files with the revisions they were based on

To see differences with other revision, you must use -D or -r option.
For example to compare with newest revisions:

Code:

cvs diff --brief -D now

Another way to see which files would be updated, without actually updating them, is to use -n global option, meaning 'Do not change any files'.
Example:

Code:

cvs -n update -dP

Now, back to work.

How to allow ordinary (not root) users to mount/unmount everything.

dr_agon — Mon, 06 Oct 2008 18:50:48 GMT

applies to: linux, debian based (Mepis), mount, sudo

While testing many things often I have to mount/unmount resources, and I became tired of using 'su' and supplying password.

The easy but not very secure solution is to allow some (or all) users to execute mount/unmount as root without being prompted for password by adding them to /etc/sudoers file like this:

Code:

%users  ALL=NOPASSWD: /bin/mount, /bin/umount

'users' is a group name.
Now I can use

Code:

$ sudo mount ......

without password.

Be careful about possible security risk!
For more informations: man sudoers.

recover from HP Backup&Recovery partition mess

dr_agon — Sat, 07 Jun 2008 13:13:04 GMT

After initial OS installation (WindowsXP) on a HP laptop I created a set of recovery DVDs. The user installed more and more unnecessary programs and remnants after applications removal started to build up, so I decided to revert it to the fresh state after the first run. The important thing is, that after creating the backup DVDs I resized the first (system) partition, created another for user files, and removed the HP Recovery partition, because I decided to keep my backups elsewhere. I didn't change the partitions before first OS installation, because there was a factory preinstalled WindowsXP image on the hard disk.
So, I had a hard disk with 2 partitions: system and user.

I booted from the recovery DVD and chose to restore "entire disk", because other options were to leave existing files on disk, and I wanted to get rid of them. There was no option for choosing the place to which restore the disk contents, neither an option to restore only one partition. I spent a while considering this, but decided to give it a try, thinking that Windows usually named a partition - "disk" ("Disk C: full" etc.). I hoped it will restore the entire backed-up contents to the first partition. I had a full, fresh backup of my disk anyway.
Well, it did exactly that. But... also something more!

It recreated the HP Restore partition, but in wrong way! It shrunk the NTFS filesystem inside the first partition (system), shrunk the second partition (user) without updating the NTFS information inside it, and put the contents of the restore partition at the end of the disk!

The result made me laugh:
I had:

the first (system) partition in the original size, but with shrunken filesystem,
the second partition (user) shrunken in partition table, but unusable because NTFS was not adjusted to smaller partition,
and the third partition (recovery) present as a ghost at the end of the disk, without proper partition table entry.

The system was bootable, but I could not access the user partition files from WindowsXP.

I had to straighten things up.
Using my favourite SystemRescueCd, which I found most helpful many times, I did the following modifications:

testdisk found the recovery partition and restored it putting necessary information into the partition table, and also updated the partition table entry for the system partition, shrinking it to the size of filesystem - not very wise, it left unpartitioned space.
I tried ntfsfix on the user partition, which probably fixed the problem with different size of filesystem and partition (although I'm not sure, maybe testdisk did it)
To resize the system partition to fill the free space I had to use Ranish Partition Manager, because I had trouble with fdisk, parted, and gparted trying to use 255 heads geometry, while all partition table entries were using 240 heads.
ntfsresize adjusted the NTFS on the system partition without any problems.

I have now system, user and recovery partition working.

Many thanks to SystemRescueCd team and NTFS utils team.

Intro

dr_agon — Sat, 07 Jun 2008 10:11:48 GMT

Hello everybody!
I sometimes solve a problem, that may be of interest to others. It usually takes a lot of internet search, and taking information from many sources. That's why it is not easy to find the most appropriate place to put a feedback or post a "thank-you". I think the easiest way will be to start this blog. If it helps someone else, I will be happy.
-----------
http://www.leksoft.com.pl/digits-wc.gif http://counter.digits.com/wc/-d/4/nchd