Palin's technical blog

Fixing click to focus problem on Gimp-X11 on MacOS X

2011-02-09T13:37:00.000+01:00

If you try to use gimp on MacOS X, you'll find that all the clicks in the tool window needs to be repeated twice, once to select the window and once to select the tool.

If you don't like this beaviour, I stealed from this site the solution:

1. Open a terminal
2. Write

defaults write org.x.X11 wm_click_through -bool true

3. Restart X11 if you have it open
4. Enjoy

Fix MSN certificate problem with Pidgin

2010-11-19T10:12:00.001+01:00

To make a long story shot, grab the certificate (here) and install under Pidgin.
Screenshots here.

Update: Of course the update to Pidgin has been issued and all. Just a reminder.

Come utilizzare Time Machine su share Samba (non supportato)

2010-10-15T11:49:00.000+02:00

Eccomi con la versione italiana del mini-howto. È probabile che aggiunga degli screenshot non appena rimetto le mani sul MacBookPro.

Ieri mi sono cimentato nel convincere TimeMachine a utilizzare una share di rete samba come disco di backup, dato che ho completato la configurazione del mio MSI Wind Box con dual Atom.

Così, dopo un po' di prove, controprove, smanettamenti e ricerche su Google, ho trovato questo post sul forum di macrumors, che mi ha fatto risolvere tutti i problemi.

ATTENZIONE: Questa modalità d'uso di TimeMachine non è supportata da Apple, quindi usatela a vostro rischio... eccetera.

Richiede l'uso del terminale.

Questi sono i passi salienti, con le istruzioni e alcuni miei commenti:

1. Rendere lo storage "compatibile" con TimeMachine.

Create un file vuoto nella radice dello storage, utilizzando da terminale il comando "touch .com.apple.timemachine.supported". Come farlo dipende da che tipo di accesso avete allo storage, potete farlo direttamente da linea di comando Linux, oppure montando la share ed eseguendo il comando da terminale su OSX. Notate che dato che il comando produce un file che inizia per "." il file stesso sarà nascosto sotto Linux e sotto OSX. Dipendentemente dalle configurazioni di Samba, sarà visibile o meno sotto Windows.

2. Creare l'immagine disco sparsebundle che verrà usata per il backup

Sul Mac, create un file sparsebundle che si chiami "Computername_macaddress.sparsebundle" (es. "Applecore_002500d1c6f0.sparsebundle") utilizzando l'Utility Disco. Il nome dell'immagine (non del file) si deve chiamare "Backup di Computername". La dimensione dovrebbe essere uguale alla massima dimensione supportata dalla share, personalmente consiglio di farla più piccola di qualche GB, oppure di dimensionarla sino al massimo spazio che volete che occupi il backup sulla condivisione. Impostate il tipo di filesystem uguale a quello utilizzato dal disco principale di OSX (solitamente MacOS Extended Case insensitive (Journaled)). Il formato dell'immagine dovrebbe essere "sparse bundle disk image". Create poi il file localmente.

Se necessitate di conoscere il macaddress, ricordate che vi serve quello della scheda ethernet del mac, anche se i backup li farete via AirPort. Non so che mac address inserire nel caso del Mac Air. Di solito il modo può veloce per ottenere il mac address di en0 è quello di usare il comando "ifconfig -a" (attenzione perché detto comando vi da il mac separato da due punti, quindi non fate copia/incolla alla cieca).

. Rendere l'immagine compatibile con TimeMachine

Montate l'immagine che avete appena creato in locale in Finder e posizionatevi nella sua radice usando il terminale (es. cd "/Volumes/Backup of Computername"). Lanciate il comando "touch .com.apple.timemachine.supported"

4. Espellere l'immagine disco

Espellete l'immagine dal Finder, è il modo più semplice.

5. Connettere la share in Finder, e copiarvi l'immagine disco

Utilizzate "Connessione al server" nel Finder per aprire la connessione desiderata, e salvate la password. Copiate l'immagine vuota locale appena creata sulla condivisione.

6. Verificate i permessi dell'immagine copiata sul server.

Se utilizzate la share samba su linux, assicuratevi che il file caricato abbia i permessi corretti. Per un backup di test io lancerei un "chmod 666" sull'immagine appena trasferita, dando a tutti i permessi di lettura e scrittura.

7. Utilizzate il disco su TimeMachine

Utilizzate le Preferenze di TimeMachine per selezionare il disco di backup, se avete ancora la share aperta nel Finder, dovrebbe trovare tutto.
Dopodiché, fate iniziare il backup. Buona fortuna :)

Problematiche

Se incontrate problemi, per prima cosa controllate e ricontrollate i permessi sulla share e sul filesystem sottostante, TimeMachine ha la brutta abitudine di creare dei file temporanei come root.

Se l'immagine si corrompe (spesso durante le prove iniziali a causa dei permessi, dello spazio, ecc.) potete controllarla con l'Utility Disco. Se l'Utility Disco trova e recupera degli errori, potrebbe erroneamente modificare la dimensione "virtuale" dell'immagine. In questo caso, utilizzate Utility Disco per impostare una nuova dimensione sull'immagine, e attendete il completamento dell'operazione, altrimenti Utility Disco continuerà a mostrarvi il vecchio valore e non capirete nulla di ciò che sta succedendo.

Come al solito, queste istruzioni sul mio Mac funzionano, sul vostro magari no... prendetele per quello che sono ;)

Using (unsupported) SMB shares for Time Machine

2010-10-14T14:46:00.000+02:00

Yesterday, after setting up my Samba NAS based on a dual-Atom MSI Wind Box, I fought against Time Machine, which refuses to use a shared folder as a backup disk.

So after a long struggle, I found this forum post which make everything work.

DISCLAIMER: this use of Time Machine will be unsupported by Apple, so use it at your own risk, warranty is void, and so on.

Terminal access is required.

These are the basic steps, with comments on my own:

1. Make your storage "compatible" with Time Machine.

Create an empty file in the root of the storage, using "touch .com.apple.timemachine.supported". Note that by beginning with a dot, the file will be hidden in OSX and in linux, but not in windows.

2. Create the disk image that will be used for the backup

On your Mac, create a sparse bundle image called Computername_macaddress.sparsebundle (ie, Applecore_002500d1c6f0.sparsebundle) using Disk Utility. It should have a name of "Backup of Computername". The size should be the maximum size the sparsebundle can grow to on whatever linux filesystem it is going to live on. Set the filesystem type appropriately for your Mac "MacOS Extended Case in/Sensitive (Journaled)". Image format should be "sparse bundle disk image". Then create the file.

If you need to know the macaddress, remember that you need the mac address of the ethernet port, even if you are going to backup via wireless. I don't know what to do on the macbook air. Usually the command "ifconfig -a" is enough to read the macaddress.

3. Make the disk image compatibile with Time Machine

Mount the image you just created in the finder, cd into it in the terminal (it gets mounted on /Volumes/Backup of Computername) and issue the "touch .com.apple.timemachine.supported" command.

4. Unmount the image

Ejecting it in the finder is the quickest option

5. Connect the share in the finder and copy the image there

Use the "Connect to server" in the finder to open the share, and save the password. Then copy the empty image you created in the server.

6. Review the permissions of the copied image on the server

If you are using a samba share on linux, make sure the uploaded file have the right permissions. For a test I would issue a "chmod 666" on it.

7. Make Time Machine use the share

Open Time Machine preferences, and select a disk to to a backup. You should find the disk you connected earlier. Start a backup.

Troubleshooting

If you encounter issues, try to double check image and share permissions, Time Machine has a bad habit of creating files as the root user.

If the image got corrupted, try to check it with the disk utility. If the disk utility finds any error and correct them, it may change the disk image size. If it's the case, simply set a new image size using disk utility itself and wait for the operation completion.

As usual, this post instructions are at "works for me"(tm) status. :)

About Mandriva

2010-09-21T09:51:00.002+02:00

This post is about my own opinion on the whole Mandriva issue.
(for the Italian version, keep scrolling)

The Fork

I think that putting up a forked project for the "community" edition of Mandriva (Mageia) is good. This should become what Fedora is for Redhat, a community-driven distribution, high on Mandriva standards and willfull to embrace new technologies in Cooker.

The Company

The company should leverage Mageia by producing a derived distribution from the stable version, including former One live editions, and selling powerpacks. From the solid base, they should expand and integrate the Enterprise Editions

The Community

Having both Mageia and Mandriva, the community should try to integrate both distribution to work with binary packages in the contrib, plf and mib repositories, so you should have binary compatibility in the stable Mageia version and in Mandriva. Projects like Codeina should go on and simply just work.

The Future

We will see if the sinergy will work correctly as I hope. In the meantime, having a backup project like Mageia, will easy the transition for many users, including myself. For now, I'm postponing the migration to CentOS as it seemed so obvious to do it just a few days ago.

Italian

Il fork

Credo che mettere in piedi un progetto fork per la comminity edition di Mandriva (Mageia) è una buona cosa. Dovrebbe diventare ciò che Fedora è per Redhat, una distribuzione incentrata sulla comunità, che mantiene gli standard di Mandriva per quanto riguarda gli obiettivi e che non oppone resistenza all'introduzione di nuove tecnologie in Cooker.

L'azienza

L'azienda dovrebbe utilizzare al meglio il materiale prodotto da Mageia, creando una distribuzione derivata dalla versione stabile di quest'ultima, compresa la versione live One e vendere la versione Powerpack. Dalla solida base della versione stabile, dovrebbe espandere e integrare la Enterprise Edition.

La comunità

Avendo a disposizione Mageia e Mandriva, la comunità dovrebbe integrare gli sforzi per consentire a entrambe di lavorare con i pacchetti binari dei repository contrib, plf e mib; in modo da avere la compatibilità binaria tra la versione stabile di Mageia e Mandriva. I progetti come Codeina dovrebbero semplicemente funzionare, come prima.

Il futuro

Vedremo se le sinergie funzioneranno come mi aspetto. Nel frattempo, avere una seconda soluzione grazie a Megia, faciliterà il momento di transizione per molti utenti, incluso me. Per il momento, sto ritardando la migrazione a CentOS, che sembrava la cosa migliore da fare solo alcuni giorni fa.

File versioning

2010-09-20T15:46:00.000+02:00

I wrote a small script to do file-level backup/versioning of a single file or group of files as a whole, without resorting to anything beyond basic cp, tar and rm.

Basically it works like this:

$ versionbackup.sh -h
versionbackup.sh -h -v -b backups | -r retdays [-d dateargs] [-m middlechar] [-p prefixchar]
    [-r retention] [-s suffixchar] [-t tarball] file ...

-b numbaks    sets the number of backups to keep (copies retention)
-d dateformat a string to be passed verbatim to the date command. Beware!
                in this case, if you omit time in a resolution high enough
                you could end _overwriting_ files! Example, if you rotate
                the files hourly and don't specify an hour in the date string,
                but only the day, you'll end up overwriting the same file 23
                times a day... better not to use this option if you are not
                sure. Note than in this case, -m is ignored.
-m middle     sets the middle char
-p prefix     sets the prefix char
-r daysago    keeps files newer than daysago days (time retention)
-s suffix     sets the suffix char
-t tarball    creates a tarball of the input files and rotate this instead
        for convenience, the tar is automatically .gz compressed
-v            verbose, echo every operation, including option parsing
                (useful to debug command-line options)

It should work nicely on any variant of unix (tested on linux and MacOSX).

I'll find some place to post the source and link it back here.

How to write accented (capital) letters on MACOSX

2010-08-23T09:16:00.000+02:00

It's usually not a problem for me to input accented letters directly with a 1-key press. But I'm used to capitalized accented letters correctly at the beginning of phrases, since in gnome or kde you must simply hit CAPS-LOCK and the accented letter key to get something capitalized (eg. È which is third person present for the verb to be in Italian). But MacOS does not change the capitalization this way.

So I had to watch out for alternatives: compositing keyboard. In Snow Leopard (at least) you can use ALT+8 or ALT+9 and the a vowel key to get respectively an acute or grave accent. Just before pressing the vowel key you'll get the accent alone underlined (as a prompt to press a letter key).

In addition, you can user ALT+O to generate german umlauts (eg. ü) or ALT+N to generate tilded characters (eg. ñ).

Ext3 Under OSX

2010-08-11T17:00:00.000+02:00

I recently got a replacement notebook, since my favourite MSI s271 presented quite a few disk errors on smart, and it's a while that this one has the usb nonfunctional.
After looking for a new notebook, in the 12"-13" range, I found with some disappointment that those niceties come at a premium price, and most of the time they are built on cheaper consumer-level standards.
So, after some research, I went for a refurbished MacBookPro, the newest model with a 13" screen.

Now the challenge is, reading all those external disks (or internal slices/partitions) which are unfortunately ext3 formatted.

I tried not to choose the straightforward path which leads to Paragon software, and instead go the MacFuse way. This post helps a lot and basically it's all you need: http://forum.archosfans.com/viewtopic.php?f=47&p=220020

The modification to the script is really needed only if you require r/w, which is basically not advised from the authors of fuse-ext3, my tests say they are quite ok.

Now I need a way to bypass the default users and permissions that macosx uses to be able to write/read the data both under linux and under osx without touching sudo or becoming root.

Mutt and Imap

2010-07-22T12:47:00.007+02:00

Yesterday I was having problems using the corporate webmail (yes, the one with three letters in the acronym), and lacked the chance to use a full-featured mail user agent.

So i remembered the old days :) when I used mutt and fetchmail to read my personal mail... and recalled mutt supported IMAP too. I started googling to find a tutorial and read a couple of nice pages (see below). In a pinch I was able to read imaps mailboxes in the console.
Here the very minimal stuff you need:

set mail_check=60
set timeout = 150
set spoolfile=imaps://$DOM\\$USERNAME@$IMAP_SERVER/INBOX
set folder=imaps://$DOM\\$USERNAME@$IMAP_SERVER/
set postponed="=Drafts"
set record="=Sent\ Items"
mailboxes =INBOX =INBOX/Folder1 =INBOX/Folder2 =Sent\ Items =Drafts
set header_cache="/home/$USER/tmp/mutt-cache"

Of course if you don't need to set a $DOM, you can skip it.

If you need more fancy colours, here is an extract I user from the original sample.muttrc:

#
# Color definitions
#
#
#color normal white default
color hdrdefault red default
color quoted brightblue default
color signature red default
color indicator brightyellow red
color error brightred default
color status yellow blue
color tree magenta default      # the thread tree in the index menu
color tilde magenta default
color message brightcyan default
color markers brightcyan default
color attachment brightmagenta default
color search default green      # how to hilite search patterns in the pager

color header brightred default ^(From|Subject):
color body magenta default "(ftp|http|https)://[^ ]+"   # point out URLs
color body magenta default [-a-z_0-9.]+@[-a-z_0-9.]+    # e-mail addresses
color underline brightgreen default

Here are the references I used to find IMAP information:

Mutt wiki: http://wiki.mutt.org/?MuttGuide/UseIMAP
Mutt documentation: http://mutt.sourceforge.net/imap/

Linux with no primary partitions left

2010-07-15T10:33:00.007+02:00

I got a new company PC, an Acer Aspire 5372ZG, nothing to say about it, just yet another consumer notebook with a 15.6" 16/9 gloss display, very big and quite heavy, nice keyboard if you don't mind having the home and end key on the keypad only.

The problem is, the 320GB HD is partitioned into three partitions, two of them are hidden. All are primary partitions. No primary partition left to install linux. So I did my homework and discovered that unlike the lilo of old, Grub is able to boot quite nicely from any filesystem it is compiled to support.

So, I created an extended partition and some logical ones with this schema:

Device Boot      Start         End      Blocks   Id  System          Mount      FS
/dev/sda1               1        1567    12586896   27  Unknown                
/dev/sda2   *        1568        1580      104422+   7  HPFS/NTFS
/dev/sda3            1581        5757    33551752+   7  HPFS/NTFS
/dev/sda4            5758       38913   266325570    5  Extended
/dev/sda5            5758        5822      522081   83  Linux        /boot      Ext3
/dev/sda6            5823        7910    16771828+  83  Linux        /          Ext4
/dev/sda7            7911        8432     4192933+  82  Linux swap   swap       swap
/dev/sda8            8433       38913   244838601   83  Linux        /mnt/data  Ext2

Of course the DATA partition is mounted under windows too, so I preferred not to have journals to reply in case of a linux unclean shutdown. It has an inode size of 128 too to be compatible with IFS tools.
I could have partitioned the system more (by creating /home for instance) but I think this setup is reasonable enough.

Ext2 under Windows

2009-10-12T22:13:00.007+02:00

I finally had the definitive data loss on my (historical) FAT32 partition that I used to share between Linux, Windows and the odd other OS (sometimes FreeBSD, sometimes OpenSolaris). So I was looking for a way to share those data without the hassle of a yesterday filesystem. I was rather disappointed with the status of NTFS under linux, both native and under FUSE, not to speak about my mascotte ZFS.
Finally, I found this Ext2 driver with finely works under Windows, reaching near NTFS native speed, so I adopted it on the pc I switch OSes more, currently the Asus EEEPC 1101HA (yes, I finally got one).
The driver is nice, if you don't mind using your favourite ext3 filesystem in compatibility mode, which only means under Windows there's no journal, and you must bear the casual fsck when windows hangs, and the fact that you can't access your filesystem under windows if it has a dirty status on linux (the journal needs to be replayed). Of course you can simply make this filesystem a simple ext2 and get on with it.
By the way, you find everything here: http://www.fs-driver.org/

Note: remember to create the ext2/ext3 filesystem with the option -I 128, which means an inode size of 128, most moder distribution set this value to 256 which is not compatible with these tools.

Update: It works under Windows 7 too with some quirks. The first is that the installation requires Windows Vista compatibility mode (not an issue for me). The second is that the drive letter utility seems not to stick with the configuration done upon reboot. I'll investigate this further.

Firefox 3.5

2009-07-02T16:19:00.003+02:00

As usual, I upgrade the pc I use at work before anything else: I use it for more time and I usually have to fix issues quickly. It's a Mandriva 2001 Spring, of course, and firefox has been installed in /opt using the tarball from the Mozilla Foundation.

Apart of the usual addons that get disabled, everything seems fine. It seems pretty responsive too, in spite of the low specs of the machine (Intel P4 2GHz, 768 MB of Ram, slow ide disk). A couple of crashes hit the usability, but thery a mostly related with plugins, maybe they should be updated too.

I don't use many of the new features, so I don't comment on them.

Resume sftp transfers

2009-06-25T08:45:00.002+02:00

Of course there should be some way to resume transfers done via sftp.

Technically the way presented here is not sftp compliant (uses curl or, better, rsync --inplace).

But it's a way to start.

Note that I don't know if sftp subsystem supports resuming natively...

Heartbeat and two-nodes clusters

2009-06-23T11:34:00.005+02:00

Heartbeat (from http://www.linux-ha.org/heartbeat/) is a nice software to manage linux high availability clusters. When you are in need to use a two-nodes cluster you want to prevent a situation known as "split-brain": when a node does not communicate with the other but neither two know which is the failing node.

This can lead to having the cluster drop all the services or having both nodes with the services on (in case of shared storage, this will lead to corruption unless the filesystem is clustered and supports multiple writes from multiple nodes). To make a long story short, split-brain is bad.

To counter this effect, you can use a ping host. Ping hosts are fake cluster members used by heartbeat to use the network connectivity as a mean to break ties in cluster node weight computations: if a node pings a ping host and the other doesn't, the hosts which pings is the winner. To make this configuration successful, you need an additional link between the nodes. This additional link could be serial or another ethernet.

To use ping hosts, this is the bare minimum configuration:

ping 10.10.10.10 # insert the ip of the ping host.
respawn hacluster /usr/lib64/heartbeat/ipfail # use /usr/lib on 32 bit hosts
apiauth ipfail gid=haclient uid=hacluster # this is for process security

But which ping host to chose? Use a reliable ping host; this can be the default gateway (if it's a router which has a very high uptime) or a local layer 3 switch. Don't use a server o a workstation!

Use hearbeat reload to configure the ping host without stopping the cluster and lose resources.

Linux first support for usb 3.0

2009-06-12T08:49:00.002+02:00

The interesting news is that Linux could be the first mainstream OS to support Usb 3.0. Really, the support is already there if you mind building your own kernel.

The good news is that finally some female hacker comes to the surface. I felt rather fed of seeing Linus or Andrew Norton faces on articles (not minding Richard Stallman's). :)

Flashblock in Opera

2009-06-05T10:26:00.002+02:00

I tried successfully the following operations to activate Flashblock in Opera (it worked with Opera 9.64 on Linux):

1. Download this file http://lexi.ucoz.ru/userjs/FlashBlock.js and put it in a directory of your choice
2. Go to Tools|Preferences|Advanced|Content|Javascript Options|User Javascript Files|Choose and choose the directory where you downloaded the javascript file.
3. Try it ;)

Note to Italian Opera users: you must use Strumenti|Preferenze (CTRL+F12)|Avanzate|Contenuti|Impostazioni per Javascript|Cartella dei file Javascript dell'utente|

Comparison of Enterprise Unix OS

2009-05-29T09:03:00.003+02:00

Here's a nice feature-to-feature comparison and quick start of major Enterprise OS. IBM Aix, HPUX and Sun Solaris are covered.

It's worth noting that only Sun Solaris is currently fully supported in x86 environment and AIX lacks any type of support for architectures different from POWER.

The article comes from IBM so maybe is a bit biased.

Get a list of most populated tables in mysql

2009-05-20T10:41:00.002+02:00

If you are a sysadmin and do not have a fellow mysql dba, you should learn some tricks to look for big tables or tables used very much.

This query (and modified versions of it) accomplishes the task:

mysql> SELECT table_name, table_schema, engine, table_rows, data_length FROM information_schema.tables where table_schema not in ('information_schema', 'mysql') order by table_rows desc limit 20;

This searches for every table which is not part of 'information_schema' and 'mysql' (that is every user database).

The result is similar to this one (names and schemas removed to protect the innocents ;))


+-----------------------+--------------+--------+------------+-------------+
| table_name            | table_schema | engine | table_rows | data_length |
+-----------------------+--------------+--------+------------+-------------+
|                       |              | MyISAM |    2828398 |   162470312 | 
|                       |              | MyISAM |    1609232 |   131429376 |
|                       |              | MyISAM |    1195717 |   218651496 |
|                       |              | InnoDB |     574491 |    43581440 |
|                       |              | MyISAM |     565128 |    10737432 |
|                       |              | MyISAM |     480290 |    26981100 |
|                       |              | MyISAM |     380521 |    22509908 |
|                       |              | MyISAM |     280128 |     8536752 |
|                       |              | MyISAM |     227324 |    18714388 |
|                       |              | MyISAM |     210842 |    17857596 |
|                       |              | InnoDB |     201526 |    45694976 |
|                       |              | MyISAM |     177225 |    10055608 |
|                       |              | InnoDB |     123960 |     5783552 |
|                       |              | MyISAM |     119143 |     9524852 |
|                       |              | MyISAM |      99960 |    10038104 |
|                       |              | MyISAM |      97712 |     8010128 |
|                       |              | InnoDB |      94167 |     8404992 |
|                       |              | MyISAM |      93962 |    16401252 |
|                       |              | MyISAM |      91817 |     1836340 |
|                       |              | InnoDB |      84424 |     4734976 |
+-----------------------+--------------+--------+------------+-------------+
20 rows in set (0.18 sec)

Multipath and partitions on linux

2009-05-19T15:50:00.004+02:00

If you have a multipath device, you can partition it. This can have many purposes, typically to use some space differently if you have a cluster and only a LUN to use as quorum disk and for data.

Sometimes, you partition the multipath device with fdisk, but the device is "in use" by the kernel and partition devices (mpathXp1, mpathXp2 and so on) don't get created.

Simply use kpartx to create those devices in "userspace":

root# kpartx -a /dev/mpath/mpath0

If you are not sure, leave the -a to preview the results.

Unison on Windows

2009-05-19T13:01:00.007+02:00

Since we are moving from one place to another (still in Rome), i needed to sinchronyze files between my current linux workstation and the notebook I was given. For me, the obvious tool is unison, and I went looking for instructions to make it work on windows, as on the notebook it's installed a windows-only VPN software.

The process turned to be fairly complex, as usual due to dll/path hell which is the doom of windows machines when dealing with a no-installer software.

1. Getting Unison and prerequisites.
The unison Homepage points here: , where I downloaded the gtk-enabled windows executable.
Unison needs the gtk library, the page suggests using those provided by Pidgin. I already had Pidgin installed, so I did nothing.

2. Putting Unison where it would find gtk dll
The page also advise to append the gtk library directory to the %PATH% variable. I opted instead to move the Unison-2.27.57 Gtk+.exe executable in the GTK directory ("C:\Programmi\File comuni\GTK\2.0\bin" in my system). This is more straightforward and easier to do and lets me have more than one GTK installation if need arises. Note that you cannot register the dll to be automatically found (I don't know why, but a dialog tells me so when I use regsvr32 on libgtk-win32-2.0-0.dll.

3. Getting ssh.exe plink wrapper
From this page [http://www.wischik.com/lu/programmer/unison-ssh.html] I downloaded the plink wrapper (I googled for unison windows ssh and it was the first hit). I put in the same folder that the unison executable. It has a bug and it does not work with a plain password (I planned to use a private key anyway)

4. Generating private key to use with plink
It's easy, use puttygen (you should have it if you installed putty with the standard windows installer, otherwise grab it here). Note that the Putty links points to the most current version of Putty, but it may change, so peek at the download page for updates.
As usual, generate a new key (I prefer to generate a fresh key for automated task, it's easier to revoke if it gets compromised as it does not have usually a passphrase) and publish the public key on the linux box which will act as a "unison server" in the authorized_keys file (usually ~/.ssh/authorized_keys, change the permissions to rw------- before doing anything else if you create a new file)

5. Configuring the key in the unison profile
Using the ssh plink wrapper, you need to modify the unison profile file. Create a new profile, select the local directory to configure, select the remote directory (using ssh), insert the host and the user that you granted access by publishing the public key.
Then locate the profile file (in C:\Documents and Settings\\.unison on my windows system) and add these lines:

sshcmd=ssh.exe
sshargs=-i path\to\privatekey_file_on_windows.ppk
ignorecase=true

you can add also perms=0 since most of the times the permissions will be totally screwed from a unix fs to ntfs (god forbids using FAT32 nowadays!)
Note: the path you specify with the sshargs option must not contain spaces.

6. Run and test
Now run Unison and verify that it detects differences in the files and that it makes good suggestions on which file is newer on which system.

Troubleshooting
The only thing you can do wrong is the public key stuff, check with cmd, cd to the directory where you have copied ssh.exe and try ssh manually:

C:\somedir> ssh -l username -i path\to\privatekey_file_on_windows.ppk hostname

and see if you can login.
Otherwise you screwed something up; check the authorized_keys name, permissions and position on the filesystem (and if it contains the right public key).

Updated Mandriva 2009 to spring

2009-05-04T09:03:00.002+02:00

The update did complete successfully, but only after manual intervention.

The update process looped on trying to install mplayer (mandriva version) while it conflicted with mplayer plf version. I had to remove the mplayer plf rpm manually from the console before the script went on.

When the notebook rebooted for the new kernel, X did not start; but mandriva offered a text mode dialog to reconfigure (and test) it.

I still have to see a linux distro upgrade flawlessly; but this time mandriva quite did it, if I did not tinker with plf, everything should go on in an acceptable way.

So the vote on the update process is 7/10.

Upgrading Mandriva 2009 to spring

2009-05-03T21:50:00.003+02:00

Tonight I started the automatic (i.e. graphical) upgrade of my notebook to mandriva 2009.1 (a.k.a 2009 Spring)...

More info to follow after upgrade.

AIX dynamic linking library

2009-04-28T12:43:00.007+02:00

When updating libraries on AIX systems, you could get this message:

[...] Cannot open or remove a file containing a running program.

even if you are updating a library file which is not in use.

Apart of verifying that the file is not in use for real, you should cleanup the dynamic library linker cache, with the command:

> slibclean

And try again...