First of all let’s see how the Sudoku problem was described at class: we have a table with 9 rows and 9 columns;

• 1. Each field [i, j] (where i=1..9 and j=1..9) has at least one value (between 1 and 9).
• 2. Each field [i, j] (where i=1..9 and j=1..9) doesn’t have more than one value.
• 3. There isn’t any repeated value in any row, column or 3×3 group.
• 4. Some of the fields have a predefined value.

Now to implement this in code, first of all I needed a Python module implementing SAT solving. A quick search in Debian’s repositories gave me python-logilab-constraint, which I’ve found to be quite nice to use, even though it could definitely take some speed improvements.

Conditions 1 and 2 aren’t a problem at all, as logilab.constraint can be used quite naturally [0]. We just define a variable for each field (eg., x11 to x99, where the first number is the row and the second number is the column) and the domain in which they operate (integer value from 1 to 9):

values = range(1, 10) # [1..9]
variables = ["x%d%d" % (i, j) for j in values for i in values]
domains = {}

for variable in variables:
	domains[variable] = fd.FiniteDomain(values)

The 4th rule is also straightforward, we just need to hardcode the values. If we have a bidimensional list sudoku containing the initial numbers and None in all empty fields, we add each of them as a constraint:

constraints = []
for i, row in enumerate(sudoku):
	for j, field in enumerate(row):
		if field is None:
			continue
		variable = "x%d%d" % (i+1, j+1)
		constraints.append(fd.make_expression((variable,),
			"%s == %d" % (variable, field)))

Now only rule 3 remains; here we basically have to set up three more groups of constraints: one for rows, one for columns and one for the 3×3 groups. My initial implementation checked each row/column/group at once; for example, for the first row «x11 != x12 != x13 != … != x19», for the first column «x11 != x21 != … != x91», etc. However, this proved to be extremely slow, and after checking the «Performance considerations» section of Logilab Constraint’s documentation I split up the row and column conditions [1] to lots of smaller conditions, as in: «x11 != x12», «x11 != x13», «x11 != x14», etc. I also moved the constraints for the initial numbers to the top (I had them at the end of the constraints list before), as they are the simplest ones. With those changes resolution time changed from several minutes to some tenths of a second.

And this is it. After all constraints have been added, we just need to run the solver:

repository = Repository(variables, domains, constraints)
solutions = Solver().solve(repository)

The complete code is available via Bazaar at lp:~rainct/+junk/sudoku-sat. Being completely new to the logilab.constraints module, or implementing any such stuff at all, it took me around half an hour to write this, which shows how SAT makes such sort of problems really straightforward.

[0] Using logilab.constraint it’s possible to assign arbitrary Python data to variables (here we just give each an integer, but variables could also take tuples or whatever else). When this problem was presented at class using pure propositional logic it was a bit more cumbersome, as we couldn’t just say “there’s a variable x11 with domain [1..9]“. For instance, rule 1 was «(p111 | p112 | p113 | … | p119) and (p121 | … | p129) …», where “p111″ would be True if field [1,1] is supposed to contain a one, “p112″ is True if it’s supposed to contain a two, etc.
[1] I didn’t bother also splitting up he 3×3 group constraints since the other two changes already gave me enough of a speedup; changing that may squeeze a few msecs more out of it.
P.S.: If you’d like a more formal explanation of this, a search on Google found this paper: A SAT-based Sudoku Solver.

Related posts:

1. You no longer have an excuse not to look at Python Snippets!

One comment
© Siegfried-Angel Gevatter Pujals, 2010. | Permalink | License | Post tags: logics, python

Last Thursday there was a talk on Acire and a snippet creation party (as part of the Ubuntu Opportunistic Developer Week) and someone mentioned it would be nice to have a web interface. So, yesterday I took a couple hours and hacked together a quick script to convert the snippets to HTML, resulting in this snippets directory which you can use to look up stuff when you haven’t Acire at hand or to give links to snippets to other people!

The new web interface to python-snippets!

And now that you’ve read this post, maybe you would like to delve into Clutter, update your application to work with Application Indicator, see how easy it is to get data from Zeitgeist or refresh the knowledge on lists you gained at Rick Spencer’s Python talk?

The code for the snippets generator is available here. Thanks go to Mako Templates, Pygments and Chardet for having made it dead easy to create this page! And, of course, patches are welcome.

Related posts:

1. Sudoku solving with Python and SAT
2. Introducing espeak-gui
3. UDS 2009

One comment
© Siegfried-Angel Gevatter Pujals, 2010. | Permalink | License | Post tags: opportunistic development, python, Ubuntu, website

Some weeks ago I implemented a new feature in Zeitgeist and I figured I’d drop some lines about it. I’m talking about a data-source registry.

You may be wondering, «what the hell is that even, a data-source?». So Zeitgeist is a database, a global event log, but it doesn’t do any magic indexing or monitoring by itself;  the information it logs needs to come from somewhere – be it applications sending it to Zeitgeist themselves, daemons, plugins for applications, etc. Any such entity is called a «data-source».

Having a register of all data-sources interacting with Zeitgeist provides some benefits, like for example:

Prototype of a data-source management user-interface

• Having a list of them. Lists are nice :).
• Being able to disable data-sources from a centralized place instead of requiring each to write its own preferences dialogue.
• Being able to set up a blacklist rule considering which data-source the information comes from as a condition (not implemented yet).

More details

When they register, data-sources will need to provide the following information: an unique ID, a name (may be localized), a description (may be localized) and optionally a template specifying what sort of events it logs. Additionally, the last timestamp the data-source was seen by Zeitgeist, whether it is running right now and whether it is enabled will also be available.

It is important to note that registration is not compulsory; while it is highly encouraged for data-sources to use it, is it still possible to anonymously insert information into Zeitgeist (for example from a shell script). The event template is also only informational, and will not be enforced by Zeitgeist at this time.

Avoiding duplicates from GTK Recently Used

You may know that zeitgeist-datahub provides basic support for applications which have no direct Zeitgeist support but do use GTK’s RecentManager, which is not as detailed as we would like, but it is better than nothing. However, until now we had a problem: when applications had support for both, GTK’s Recently Used and Zeitgeist (be it native or as an extension), it was possible for duplicate events to be inserted or other sorts of conflicts between both data-sources. Now that we have the information from the registry available, we’ve been able to solve this modifying our Recently Used data-source to ignore any events concerning applications which already have another data-source logging the same types of events.
 

Related posts:

1. Zeitgeist is out!
2. Introduction to Zeitgeist 0.2’s API
3. Zeitgeist Hackfest (II)

One comment
© Siegfried-Angel Gevatter Pujals, 2010. | Permalink | License | Post tags: gnome, Programari lliure, zeitgeist

I think the time has come that I make a confession. You may be wondering why I haven’t spent much time with you those last months, and you have the right to know. The case is, I have another one. She’s called Debian.

No, it’s not because of you. It’s also not because of your parents; while they aren’t perfect, they are great in many ways and I certainly don’t want to run away from then. Just wait and let me explain this. You surely remember that I spent a week with her some time ago? Well, now that I want to join Debian’s family, I decided that after all that wasn’t much time to get to know here, so I gave her another chance, and with this second experience I fell in love with her unstable character. Ubuntu, you are really extraordinary, and I’m still telling anyone who wants to listen about your wonders, but you just can’t appease my desire for trying out new things as well as Debian does.

Now, this doesn’t mean we won’t see each other anymore. I’ll still take you with me whenever I travel, and I’ll continue helping you with the household. In fact, all this wouldn’t signify a big change at all, if it wasn’t that at the same time I also decided to refocus most of my efforts on raising children, which is what really reduced the time I spent helping you.

So, while currently I’m not seeing you as much as before, I hope you all the best, and I’ll still try to help you advance, be it indirectly (1, 2) or, while probably to a lesser degree, continuing with direct contributions.

By the way, hello Planet Debian!

8 comments
© Siegfried-Angel Gevatter Pujals, 2010. | Permalink | License | Post tags: Debian, Programari lliure, Ubuntu

By the way, if you’re coming too and are interested in Zeitgeist, don’t forget to poke me (or Seif)!

Related posts:

1. Zeitgeist since UDS
2. Zeitgeist is out!
3. Introduction to Zeitgeist 0.2’s API

No comments
© Siegfried-Angel Gevatter Pujals, 2010. | Permalink | License | Post tags: fosdem, gsoc2009, Programari lliure, zeitgeist

While several sources have already propagated the good news, what doesn’t seem to be so widely known is how easy it is to get the Activity Journal running on Ubuntu. Because it is!

Installation on Ubuntu Karmic or Lucid

sudo add-apt-repository ppa:zeitgeist/ppa
sudo aptitude update
sudo aptitude install gnome-activity-journal

Now go to Applications -> Utility -> Activity Journal and enjoy the sweetness!

Installation on other systems

Our Personal Package Archive also has packages for Ubuntu Jaunty which you can add the usual way (by adding it to System -> Administration -> Software Sources or editing your sources.list), and the same packages as for Ubuntu work for Debian Sid (to which Zeitgeist 0.3.2 and the GNOME Activity Journal will be uploaded shortly).

There may be packages for other distributions available; if you can’t find any for yours, you can do the installation by hand:

[For Zeitgeist] – Build dependencies: intltool

$ wget http://launchpad.net/zeitgeist/0.3/0.3.2/+download/zeitgeist-0.3.2.tar.gz
$ tar -xzvf zeitgeist-0.3.2.tar.gz
$ cd zeitgeist-0.3.2
$ ./configure
$ make
# make install

[For the GNOME Activity Journal] – Build dependencies: Python (2.5+), Python-DistUtils-Extra, intltool

$ wget http://launchpad.net/gnome-activity-journal/0.3/0.3.2/+download/gnome-activity-journal-0.3.2.tar.gz
$ tar -xzvf gnome-activity-journal-0.3.2.tar.gz
$ cd gnome-activity-journal-0.3.2
$ python setup.py build
# python setup.py install

But, what is the GNOME Activity Journal?

The GNOME Activity Journal is a tool for easily browsing and finding files, contacts and other resources on your computer. Using Zeitgeist, it keeps a chronological journal of your activity and supports tagging and bookmarking (using the new Tracker 0.7) and establishing relationships between resources.

While this first release only supports basic browsing of file activities, the underlying infrastructure can do much more and you can expect the missing functionality to become available in future releases.

Related posts:

1. Another GNOME Shell post: what do you think about the window list?
2. GNOME Do – 0.8 Alpha and Docky
3. Updated Voxforge packages in Ubuntu

7 comments
© Siegfried-Angel Gevatter Pujals, 2010. | Permalink | License | Post tags: Debian, gnome, Programari lliure, Ubuntu, zeitgeist

The main changes are fixing a crash bug and… internationalization support. So, if you like this project but you’re not a coder, now you’ve got a way to contribute: translating it into your language (come on, it only has a few strings! :)).

The new version is already in my PPA: espeak-gui_0.2-0ubuntu1~ppa1_all.deb. A tarball is also available, and the installation instructions are the following:

$ wget -c http://launchpad.net/espeak-gui/trunk/0.2/+download/espeak-gui-0.2.tar.gz
$ tar -xzvf espeak-gui-0.2.tar.gz; cd espeak-gui-0.2
$ python setup.py build
$ sudo python setup.py install

In case you didn’t already have the previous version, you’ll also need to install the python-espeak module, for which there are also .deb’s (i386, amd64) or the manual installation option like explained in my previous post:

$ bzr get lp:python-espeak
$ cd python-espeak
$ python setup.py build
$ sudo python setup.py install

Here are the contents of the NEWS file:

2010-01-02: Version 0.2
-----------------------

 - Fixed a crash happening when there was no voice language set in
   gconf (LP: 502253). [Reported by Rugby471]
 - Added a "Plain text" filter to the Open/Save dialogues.
 - Added internationalization support.
 - Added a manpage and a gconf schemas file.
 - A couple little user interface improvements.

Translations: Catalan.

Related posts:

1. Introducing espeak-gui
2. GNOME Activity Journal, and installing it on Ubuntu
3. Here is Zeitgeist 0.2.1!

29 comments
© Siegfried-Angel Gevatter Pujals, 2010. | Permalink | License | Post tags: gnome, gtk, Programari lliure, python, speech-synthesis, Ubuntu

Why, when, who?

The project started almost a year ago when, out of curiosity on what writing Python bindings for C/C++ libraries is like, I started python-espeak, bindings for the espeak speech synthesizer (which, by the way, comes installed by default on Ubuntu and many other distributions).

It turned out that writing bindings isn’t as funny as I thought so they haven’t advanced much since then, but the basic functionality was there and I felt the need for some application using those bindings, so that’s how I started espeak-gui.

Some months later, in true Free Software spirit, someone else -interested in an application like this for personal usage- found my code on Launchpad and got in touch with me, providing me with quite some nice patches. Thank you, Joe Burmeister! However, I didn’t do any more work on it, as I’ve been busy with other projects (eg. Zeitgeist), and development stalled there.

Now, another half a year later, I’ve finally got back to this and decided it’s time I push it out into the wild. So, after cleaning it up a bit more and implementing some new feature, here you have espeak-gui!

(The bindings aren’t really encouraged for widespread usage at this point and I’ll probably end up rewriting them using SWIG or something else; however, if you’re interested in using them please get in touch with me).

Where do I get it?

If you’re using Ubuntu or Debian, you can find a packages for python-espeak and espeak-gui in my PPA.

For users of other distributions, you can install them manually after installing the needed dependencies (most importantly, libespeak-dev).

For the Python bindings for espeak:

$ bzr get lp:python-espeak
$ cd python-espeak
$ python setup.py build
# python setup.py install

And for the GUI:

$ wget -c http://launchpad.net/espeak-gui/trunk/0.1/+download/espeak-gui-0.1.tar.gz
$ tar -xzvf espeak-gui-0.1.tar.gz
$ cd espeak-gui-0.1
# python setup.py install

[Update: A new version is out, see espeak-gui 0.2 for details.]

I have several ideas on how to continue improving it and I think I’ll slowly continue doing so (or maybe not so slowly if I get positive feedback on this :)). Also, patches are always welcome!

Using it

Once installed, you’ll find espeak-gui under Applications -> Sound and Video (maybe Accessibility would be a better place?), or you can run it from the command line like this:

espeak [ <file 1> <file 2> ... ]

Related posts:

1. espeak-gui 0.2
2. GNOME Activity Journal, and installing it on Ubuntu
3. Introduction to Zeitgeist 0.2’s API

14 comments
© Siegfried-Angel Gevatter Pujals, 2010. | Permalink | License | Post tags: gnome, gtk, Programari lliure, python, speech-synthesis, Ubuntu

But it’s not to advertise this game that I’m writing about it, but because I want to write down some solution to the “no sound problem” which you’re likely to find if you try it out, and so that I can remember about the fixes in the future. So, here’s the different things you can try; if one of them fails just go to the next one, and let’s hope that at least the last one does the trick!

1. Install the “oss-compat” package.
As easy as that! Although this alone has never fixed the problem with me, it often does together with solution 2 (described just below), and maybe in your case it may be all you’re missing.

2. A bit of command-line magic
The next one, which seems to be the most successful option, is trying to execute the following line before running Enemy Territory:

echo "et.x86 0 0 direct" | sudo tee /proc/asound/card0/pcm0p/oss

If after doing this the sound works, you can make the change permanent by adding that line to your /etc/rc.local/, for example by running:

sudo sed -i "s:^exit 0:echo 'et.x86 0 0 direct' > /proc/asound/card0/pcm0p/oss\nexit 0:" /etc/rc.local

3. Let’s call the artillery

Hopefully you’ll be enjoying an Enemy Territory game with the sound of explosions and shots in your ears, but if the previous solutions didn’t do the trick do not despair, there’s still hope. I’ve found a different solution in this thread and it consists in the following:

a) Download this script: wolfsp-sdl-sound.sh (original download link, alternative download link), created by Pyry Haulos.

b) Open it with a text editor and customize it to match your configuration: change the line «GAME_PATH=”/home/games/enemy-territory”» to point to the location of your Enemy Territory installation (most likely «GAME_PATH=”/usr/local/games/enemy-territor”»); in case your directory with the game files isn’t called «enemy-territory» like in the example you’ll also need to adapt the GAME_DIR variable.

c) Save this file somewhere on your computer, make it executable if needed (you can do this by right clicking on it, selecting “Properties” and under the “Permissions” tab of the dialogue that will show up checking the box to allow execution of the file). Done, now whenever you want to run the game just start it by running this file.

I hope this is useful for someone out there :).

Related posts:

1. No sound on HP Pavilion dv6-1110ss
2. Lock the CD tray when the laptop’s lid is closed
3. Getting your scanner to work with Ubuntu (gt68xx)

No comments
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: Debian, Jocs, Ubuntu

Related posts:

1. Arreglant paquets que no compilen: necpp

No comments
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: Programari lliure, Xorrades

From the engine side we are implementing a new database and API which does a clean cut between what is event logging and what is repository functionality. What this means for Zeitgeist is that we won’t do annotations anymore and that the basic information Zeitgeist has about items will be about those at the time of the event – for anything else there is Tracker (I tried out the new version from git yesterday and it really rocks, it’s nothing like 0.6).

We are also implementing focus tracking and the algorithms for the fancy stuff (contextual relevancy over time, focus time of documents/applications, etc).

Blackboards saved us a lot of time. In the image: Seif Lotfy, Ivan Frade (Tracker) and Markus Korn.

The User Interface team also has some quite nice design ideas. For the non-developers out there, there’ll be something (alpha-quality!) to try out soon.

Related posts:

1. Zeitgeist Hackfest
2. Zeitgeist is out!
3. Zeitgeist since UDS

2 comments
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: gnome, Programari lliure, zeitgeist

Just a quick note to mention that tomorrow I’m leaving for Bolzano for a week, for the Zeitgeist Hackfest. I’m really looking forward to meeting all the team there!

My initial idea for this post was to write about some of the upcoming changes in the engine, as we’ve been discussing quite some improvements lately, but I’ll better leave that for after the hackfest as we’ll have a chance to go over all them again and clarify any missing chunks :). For now I’ll just say that the API will probably change quite a lot, and while I’m at it leave you with a link to a recent Status Report sent to GNOME (there’s also one for GNOME Shell, in case you care).

Related posts:

1. Zeitgeist Hackfest (II)
2. Introduction to Zeitgeist 0.2’s API
3. Zeitgeist Data-Source Registry

3 comments
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: gnome, Programari lliure, zeitgeist

1. Comprovació de l’estat

Ara no perquè ja he arreglat el paquet, però si miravem a https://launchpad.net/ubuntu/+source/necpp veiem que la última versió del paquet era l’afectada, i que no hi havia informes d’error relacionats amb el problema (recordem que, si n’hi hagués, podríem trobar-hi pistes o bé veure que algú altre ja està treballant en arreglar aquest paquet). Comprovant els informes d’error a Debian (http://bugs.debian.org/necpp) tampoc trobem res, i a http://packages.debian.org/necpp comprovem que tampoc hi ha cap versió més nova que la de l’Ubuntu on ho hagin arreglat. Així doncs, tirem endavant, i obrim un informe d’error al Launchpad per indicar que hi estem treballant (exemple).

2. Obtenció del paquet font
Baixem el paquet font:

apt-get source necpp

Si ho proveu ara, obtindreu la nova versió del paquet que ja he arreglat, però podeu baixar-vos l’anterior a fi de seguir el tutorial fent això en lloc de l’apt-get source:

dget https://launchpad.net/ubuntu/+archive/primary/+files/necpp_1.3.0+cvs20090101-1ubuntu1.dsc
dpkg-source -x necpp_1.3.0+cvs20090101-1ubuntu1.dsc

Ara, comprovem que efectivament podem reproduir el problema, intentant compilar-lo:

pbuilder-karmic build necpp_1.3.0+cvs20090101-1ubuntu1.dsc

El missatge d’error que rebem és el següent:

[...]
XGetopt.cpp: In function 'int XGetopt(int, char**, const char*)':
XGetopt.cpp:170: error: invalid conversion from 'const char*' to 'char*'
make[3]: *** [XGetopt.o] Error 1
[...]

3. Solució del problema
Arribem així a la part més difícil, ja que aquesta no és sistemàtica.

El missatge d’error ens parla d’un fitxer XGetopt.cpp, el qual podem trobar al directori str/. Si mirem a la línia que menciona, la 170, ens trobem el següent:

cp = strchr(optstring, c);

Si cerquem informació sobre la funció strchr del C++, trobem que està sobrecarregada i té dues versions:

const char * strchr (const char * str, int character);
char * strchr (char * str, int character);

Veiem que si el primer paràmetre de la funció és constant, el resultat de la funció és una constant, i sí el paràmetre no és constant, el resultat tampoc ho és. Si mirem de quin tipus és la variable optstring, a la línia 133 trobem que és constant, de manera que el resultat de la funció també ho és. Si ara mirem com és la variable cp, a la qual s’està assignant la sortida de la funció, a la línia 136 veiem:

char *cp;

És a dir, no és constant, de manera que aquí tenim el problema. Si mirem la resta del codi on existeix aquesta variable, no trobem cap lloc on es modifiqui el valor de la cadena de text a la qual apunta, així que podem convertir la variable en constant sense problema. Per fer-ho, simplement cal canviar la línia afegint-hi «const» al davant, i com que si executem l’ordre

what-patch

ens diu que el paquet no utilitza cap sistema de pegats («patchless?») podem fer el canvi modificant el fitxer directament.

Per acabar, afegim una entrada nova al fitxer debian/changelog, declarant una revisió nova del paquet i explicant els canvis que hem fet. Executem l’ordre següent:

dch -i

I ha de quedar una cosa com ara:

necpp (1.3.0+cvs20090101-1ubuntu2) karmic; urgency=low

  * src/XGetopt.cpp (LP: #447454)
     - Change the definition of the "cp" variable from type "char*" to
       "const char*", to avoid an "invalid conversion" error when the
       result of "strchr(const char*, ...)" is assigned to it.

 -- Siegfried-Angel Gevatter Pujals   Sat, 24 Oct 2009 19:58:55 +0200

És important no oblidar-nos de referenciar el número de l’informe d’error que hem obert, i explicar clarament el motiu de tots els canvis.

4. Regenerant el paquet font

debuild -S

Això ens crearà un paquet font per a la nova revisió, que podem compilar tal com hem fet abans (però utilitzant el fitxer .dsc de la nova revisió, no el d’abans):

pbuilder-karmic build necpp_1.3.0+cvs20090101-1ubuntu2.dsc

Aquest cop funcionarà, així que ara només queda generar un fitxer de diferències i comprovar que només hi ha els canvis que hem fet intencionadament.

debdiff necpp_1.3.0+cvs20090101-1ubuntu1.dsc necpp_1.3.0+cvs20090101-1ubuntu2.dsc > necpp_1.3.0+cvs20090101-1ubuntu2.debdiff

En aquest cas, el contingut del fitxer ha de ser el següent:

diff -u necpp-1.3.0+cvs20090101/debian/changelog necpp-1.3.0+cvs20090101/debian/changelog
--- necpp-1.3.0+cvs20090101/debian/changelog
+++ necpp-1.3.0+cvs20090101/debian/changelog
@@ -1,3 +1,12 @@
+necpp (1.3.0+cvs20090101-1ubuntu2) karmic; urgency=low
+
+  * src/XGetopt.cpp (LP: #447454)
+     - Change the definition of the "cp" variable from type "char*" to
+       "const char*", to avoid an "invalid conversion" error when the
+       result of "strchr(const char*, ...)" is assigned to it.
+
+ -- Siegfried-Angel Gevatter Pujals   Sat, 24 Oct 2009 19:58:55 +0200
+
 necpp (1.3.0+cvs20090101-1ubuntu1) karmic; urgency=low

   * Merge from debian unstable, remaining changes from jaunty:
only in patch2:
unchanged:
--- necpp-1.3.0+cvs20090101.orig/src/XGetopt.cpp
+++ necpp-1.3.0+cvs20090101/src/XGetopt.cpp
@@ -133,7 +133,7 @@
 int XGetopt(int argc, char *argv[], const char *optstring)
 {
 	char c;
-	char *cp;
+	const char *cp;
 	static char *next = NULL;
 	if (optind == 0)
 		next = NULL;

5. Enviant els canvis a l’Ubuntu

Ara que tenim el fitxer de diferències, l’adjuntem a l’informe d’error que hem creat al principi, i subscrivim l’equip de revisors adequat a l’informe. En aquest cas, el paquet està al dipòsit universe, com podem comprovar:

$ apt-cache madison necpp
     necpp | 1.3.0+cvs20090101-1ubuntu1 | http://mirror.pacific.net.au karmic/universe Sources

Així doncs, l’equip és l’ubuntu-universe-sponsors. Ara només queda esperar que algú revisi el paquet i el pugi a l’Ubuntu, però com que som bons ciutadans també reportarem l’error a Debian tot donant-los-en la solució.

Related posts:

1. Com contribuir a l’Ubuntu – Arreglar paquets que no compilen correctament

No comments
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: motu, Programari lliure, Ubuntu

Prefaci

Inauguro la sèrie amb aquest primer apunt, on tot seguit explicaré conceptes generals en quant als “paquets que no compilen” i que fer amb ells; en successius apunts, i per tal de deixar-ho tot ben clar, explicaré els passos necessaris per a arreglar diferents paquets concrets. Depenent de la resposta que rebin aquests escrits, pot ser que més endavant enceti una sèrie nova que tracti algun altre tema relacionat amb els paquets, a partir dels coneixements adquirits amb aquesta.

Introducció

Si esteu llegint això ja deveu saber que en l’Ubuntu els programes venen preparats en forma de paquets; gestors de paquets (com ara l’Afegeix/Elimina, el Synaptic o l’Apt) us proporcionen una llista de tots els paquets disponibles i simplement esperen les vostres indicacions per tal de descarregar, insta?lar i mantenir actualitzats els paquets que us interessin. Ara bé, per increïble que pugui semblar, aquests paquets no apareixen del no-res, sinó que hi ha una munió de gent dedicant temps a crear-los i actualitzar-los.

Quan hom crea un paquet el que fa és agafar el codi font del programa (amb petites modificacions, si convé, com ara arreglar-ne algun error) i afegir-hi la informació necessària per a crear un programa. Un cop això està fet, el paquet font (source package) resultant és enviat al Launchpad (servidor d’Ubuntu), on és compilat per a les diferents arquitectures, tot creant els paquets binaris (.deb) que tots coneixem. Pot ocórrer que un paquet font no compili correctament (i no es creïn els paquets .deb) per a una o més arquitectures; en aquest cas, cal trobar quin és el problema i enviar una nova revisió del paquet al Launchpad. Del fet que un paquet no compili correctament se’n diu FTBFS («Fails to Build From Source»).

Normalment, poc abans que surti una nova versió de l’Ubuntu, s’intenta tornar a compilar tots els paques dels dipòsits oficials per tal de comprovar que encara compilin correctament. Aquesta operació acaba donant un llarg llistat de paquets que fallen i que convé arreglar. Els motius que els paquets no compilin correctament són diversos, i poden ser coses com ara: l’ús d’una nova versió més estricta d’un compilador (com passa sovint amb el gcc), canvis en les dependències de construcció del paquet (és a dir, en els paquets de que cal disposar per crear-lo, «build dependencies»; pot ser un canvi en el contingut d’algun d’aquests paquets el que provoqui el problema, que hagi canviat el nom d’algun d’ells i ara no el trobi…), etc.

La llista de paquets que no compilen correctament a la Karmic es troba aquí: Build status for Ubuntu Karmic. Passem a veure, en termes generals, que fer amb aquests paquets.

0. Preparació abans de tractar amb paquets (això només cal fer-ho el primer cop)

Abans de posar-nos mans a la obra ens cal preparar el nostre sistema, instal·lant-hi unes quantes eines.

sudo aptitude install ubuntu-dev-tools devscripts debhelper cdbs patchutils pbuilder build-essential

Fem saber el nostre nom a aquestes eines, obrint el fitxer .bashrc i afegint les dues línies següents al final (tot substituint les dades per les nostres pròpies, és clar!):

export DEBEMAIL=usuari@proveidor.com
export DEBFULLNAME="Nom Cognom"

[En cas que disposem d'una clau PGP, convé escriure el mateix nom i l'adreça de correu electrònica tal com els tenim a la clau, tot posant el comentari -si n'hi ha- entre parèntesis i a continuació del nom.]
També crearem un entorn Ubuntu mínim on podem provar de compilar els paquets sense que afectin (ni siguin afectats) pel nostre sistema. [Si no estem utilitzant la Karmic, cal insta?lar el paquet debootstrap del dipòsit -backports corresponent a la versió que utilitzem. També és recomanable agafar la última versió de l'ubuntu-dev-tools.]

sudo ln -s /usr/bin/pbuilder-dist /usr/local/bin/pbuilder-karmic
sudo chmod +x /usr/local/bin/pbuilder-karmic
pbuilder-karmic create

Haurem d’esperar una estona fins que tot el sistema base sigui insta?lat i actualitzat. Per acabar, ens hem d’assegurar que tinguem una línia deb-src per a la Karmic (encara que estiguem utilitzant la Jaunty o l’Intrepid, al ser de codi font això no té importància) al fitxer /etc/apt/sources.list i la llista de paquets actualitzada (sudo aptitude update).

1. Comprovació de l’estat de l’error: ja hi està treballant algú?

Un cop hem triat quin paquet volem arreglar, anem a https://launchpad.net/ubuntu/+source/nom_paquet, comprovem que la última versió indicada sigui la mateixa que a la llista de FTBFS, i comprovem els informes d’error per si ja n’hi ha algun sobre el problema en qüestió. Si és així, comprovem que no estigui assignat a ningú (si ho està, deixem el paquet per ell i ens en busquem algun altre, per no duplicar la feina), mirem si l’informe conté alguna informació rellevant, i ens l’assignem a nosaltres, canviant a  més l’estat a In Progress. Si no hi ha cap informe d’error sobre el FTBFS, en creem un i igualment ens l’assignem a nosaltres i el marquem com a en progrés.

Fet això, comprovem també http://packages.qa.debian.org/nom_paquet, on ens assegurem que a Debian no hi hagi ja una versió nova del paquet que solucioni l’error ni un informe d’error amb informació que ens pugui ser útil. Si fos el cas que a Debian el paquet ja ha estat arreglat, depenent de la situació en tindríem prou amb demanar que sigui sincronitzat a l’Ubuntu (veure sync requests) o bé, si el paquet d’allà té altres canvis que en aquest moment no volem (per exemple una versió amb funcionalitats noves, ja entrada en vigor la Feature Freeze), podem agafar la solució de Debian de forma isolada per aplicar-la al nostre paquet.

2. Obtenció del paquet font

Ara que hem comprovat que el problema encara no ha estat solucionat, potser hem trobat alguna informació útil, i que hem indicat que estem treballant en arreglar-lo, procedim a baixar el paquet font:

apt-get source nom_paquet

Això baixarà tres fitxers al nostre directori actual: nom_paquet_versió.orig.tar.gz, nom_paquet-versió-revisió.diff.gz i nom_paquet-versió-revisió.dsc. El primer és el codi font agafat de l’autor original (<em>upstream</em>), sense modificacions; el segon fitxer és un fitxer de diferències que conté la informació necessària per crear al paquet (que serà posada en un directori anomenat «debian» dins del codi font) i en alguns casos també canvis al codi original, i l’últim fitxer simplement serveix per verificar que els altres dos són correctes (que no ha estat modificats o corromputs pel camí). Automàticament, també descomprimirà el contingut del .tar.gz i li aplicarà els canvis del .diff.gz, de manera que trobarem un directori nom_paquet-versió on treballar.

[En el paràgraf anterior, «nom_paquet» es refereix al nom del paquet font, «versió» al número de versió que li ha posat l'autor original, i «revisió» al número de revisió del paquet, que pren la forma XubuntuY, on X és la revisió de Debian i Y la revisió d'Ubuntu -si el paquet ve directament de Debian, simplement serà X-].

Arribats a aquest punt, podem verificar que realment el paquet no compila correctament, tot intentant-ho amb:

pbuilder-karmic build nom_paquet-versió-revisió.dsc

Per a continuar, ens situarem dins del directori descomprimit:

cd nom_paquet-versió

3. Solució del problema

Ara que tenim el paquet de codi font, hem de trobar la solució al problema. Per a això ens és útil llegir el registre de l’intent de creació del paquet fallit, cercar informació als informes d’error d’Ubuntu i Debian (com ja s’ha mencionat abans) i del lloc web de l’autor original, llegir el registre de canvis del paquet (el fitxer debian/changelog dins del directori de codi font) per si hi ha referències útils, etc. Aquest constitueix el pas crític i més important, ja que si no sabem com solucionar l’error poc podrem fer.

[En cas que trobem un canvi al debian/changelog que sembli rellevant al problema actual, però no l'acabem d'entendre, preguntarem a la persona que l'hagi fet, ja sigui enviant-li un missatge de correu electrònic o cercant-lo a l'IRC (canal #ubuntu-motu o #ubuntu-devel a irc.freenode.org.]

Un cop haguem trobat com solucionar el problema, apliquem els canvis necessaris. Aquests acostumen a consistir en una modificació al fitxer debian/rules (per corregir les instruccions de creació del paquet), debian/control (per corregir les dependències de construcció) o als fitxers de codi font; les modificacions a aquests darrers cal fer-les utilitzant un sistema de pegats, en cas que el paquet ja disposi d’un (podem veure-ho amb l’ordre what-patch), o bé modificant els fitxers directament en cas contrari.

Fet això, haurem d’afegir a la informació del paquet una nova revisió. Per a això executarem «dch -i» i, a l’espai on descriure els canvis, explicarem detalladament què hem fet i perquè. També hi inclourem una referència a l’informe d’error que hem obert (o trobat) anteriorment, de la forma «(LP: #número)». Usualment el format utilitzat és d’aquest tipus:

  * debian/control: Add foo as a build dependency, because the build system invokes command bar provided by the former; this fixes a FTBFS (LP: #123456).

Com veieu, es comença posant el nom del fitxer (o els fitxers) modificat, dos punts, i a continuació s’expliquen detalladament els canvis; cada canvi es posa en una línia separada (començant per « *»).

4. Regenerant el paquet font

Hem arreglat el problema i hem documentat tot en una entrada nova al fitxer debian/changelog. Ara podem procedir a crear un paquet font per a la nova revisió; això ho farem amb l’ordre següent

debuild -S

[És estrany, però pot ser que algun paquet requereixi la insta?lació de part de les dependències de construcció per tal de crear el paquet font. Quan aquest sigui el cas, n'hi ha prou amb insta?lar les dependències mancants i repetir l'ordre.]

Si tot va bé, ara podem tornar al directori pare del codi font (<em>cd ..</em>) i ens trobarem amb dos fitxers nous: un <em>.diff.gz</em> i un <.dsc> per a la nova revisió (l’<em>.orig.tar.gz</em> no ens cal, ja que la versió de l’autor original és la mateixa que abans).

Ara podem tornar a provar de construir el paquet, i si hem fet bé la nostra feina aquest cop funcionarà (i deixarà un o més paquets binaris a ~/pbuilder/karmic_result/):

pbuilder-karmic build nom_paquet-versió-revisió_nova.dsc

Si ha funcionat, felicitats! Per acabar, crearem un fitxer de diferències amb els canvis entre la revisió actual de l’Ubuntu i la nova que hem preparat, tot fent:

debdiff nom_paquet-versió-revisió_original.dsc nom_paquet-versió-revisió_nova.dsc > nom_paquet-versió-revisió_nova.debdiff

El fitxer .debdiff l’obrirem amb un editor de text per comprovar que no s’hi hagi colat cap canvi que no haguem fet. [Si hi ha alguna cosa que no volem, haurem de desfer el canvi al directori del codi font i tornar a regenerar el paquet font, no s'hi val editar el fitxer .debdiff a mà ja que això segurament l'espatllarà.]

Enviant els canvis a l’Ubuntu

Un cop estiguem satisfets amb la nostra solució i haguem provat que funciona, adjuntarem el fitxer .debdiff a l’informe d’error que hem trobat o creat al Launchpad i només ens faltarà que un desenvolupador d’Ubuntu el comprovi i el pugi a l’Ubuntu. Per a això haurem de mirar en quin dipòsit està el paquet, fent:

apt-cache madison nom_paquet

Si la penúltima paraula és «universe» o «multiverse», subscriure’m l’equip «ubuntu-universe-sponsors» a l’informe d’error utilitzant l’opció que trobarem al Launchpad a la dreta; si, en cavi, és «main» o «restricted», subscriure’m l’equip «ubuntu-main-sponsors». Ara, tot és qüestió d’esperar que algú revisi els nostres canvis i els apliqui o bé proposi possibles millores.

Quan els nostres canvis hagin estat acceptats, també els enviarem a Debian, en cas que allà també hi siguin rellevants. Per això simplement enviarem un mail a número_de_l’informe_d’error@bugs.debian.org si ja existeix un informe d’error, o n’obrirem un de nou si no n’hi ha.

Conclusió

Si heu llegit tot això, us felicito, ja que realment heu mostrar voluntat. Com he promès al prefaci, els propers dies escriure uns quants apunts més explicant (de forma més breu) com arreglar alguns paquets particulars, tot veient solucions a diferents problemes.

Related posts:

1. Arreglant paquets que no compilen: necpp
2. GNOME Activity Journal, and installing it on Ubuntu
3. Més val prevenir… (Còpies de seguretat a Ubuntu)

One comment
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: motu, Programari lliure, Ubuntu

When it comes to non-instantly many-to-many communication, there are to big systems currently in use: forums and mailing lists. What annoys me that much about those is that both systems are actually exactly the same, except for the frontend they use.

Mailing lists send the messages to all users by mail, and allow them to answer again by mail. Additionally, they have a website (archive) where all messages can be checked, but this website is usually pretty awful and doesn’t allow answering to the messages.

Comparably, forums show all messages -and allow to answer to them- in a website which is excellent for this purpose, but the mail subscription option isn’t useful at all (often requiring that you open the website to see the content) and doesn’t allow to answer to the messages directly.

But, let’s look again at what their key function actually is. Delivering messages. It’s just, one simple purpose that both systems share. So my point is, why isn’t there a single solution adjusting to both (mail and web) workflows. Why are there only two exclusive options, which only provide one good interface and a broken attempt at the other one?

I want a system where I can visit a website and see all messages (also being able to answer to them), and then after five months decide that I don’t want to visit the website any more and subscribe to it by mail, still getting to read and answer the discussions in the same way but now from my mail client? (And while we are at it, having an option to link several email addresses to the same account, so that I can answer using any of those – like with Launchpad’s mailing lists).

In all the years I’ve been using the Internet, I haven’t seen a single solution like the one I’m describing. Sure, there are third-party websites which show discussions from mailing lists in a more readable way, and some even allow answering to them, but it isn’t quite the same. There is Google Groups, which is quite close at providing both interfaces, but the website isn’t really like a forum either and it just doesn’t feel right (also, it’s proprietary/centralized and not appropriate for usage by, for example, the Debian project).

Am I really asking for that much? I’ve considered, several times, writing such a system myself, but I just can’t believe that nobody has thought of it before (or, in case it already exists, that it isn’t in much more widespread use).

Related posts:

1. Another GNOME Shell post: what do you think about the window list?

17 comments
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: Programari lliure

[rainct, ~]$ vrms
bash: vrms: command not found

Whoops!

2 comments
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: memo, Programari lliure

What maybe isn’t that well-known is, where do those images actually come from? Easy: they are all at screenshots.debian.net and applications interested in showing them fetch them from there.

Now, I’m not telling you this because I think it’s soooo interesting, but because I’m seeing lots of packages without screenshots and… you can help create those! So, go look for some of your favourite application missing a screenshot, take a meaningful screenshot of it (in English, using PNG format, etc.) and head over to screenshots.debian.net to upload it :).

The new software centre (which can be tested in Karmic) of course also shows those screenshots, and in a much nicer way (they fade in automatically).

Related posts:

1. GPL: Not here

10 comments
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: Debian, packaging, Programari lliure, Ubuntu

Here is the full changelog:

- Added compatibility with Python version 2.5.
- Removed the Storm dependency, obtaining general performance improvements.
- Removed the need for a patched PyGTK.
- Made the GtkRecentlyUser logger more robust (fixes an infinit loop on some
systems).
- Improved performance of DeleteItems and UpdateItems.
- Fixed a problem with the contents of the EventsChanged signal.
- Fixed InsertEvents to enforce "mimetype" as a required value.
- Fixed a bug where the sorting_asc=True in FindEvents would be ignored if used
together with mode="mostused" (LP: #404947).
- Highly improved caching.
- Added a "--quit" option to zeitgeist-daemon to stop any running daemon.
- General code improvements, new test cases and other minor fixes.


Additionally, it also adds the possibility to use a list of strings (instead of just a single string) as value for the “uri” and “name” filters; if used, the condition between the different values in the list is OR (so you can do, for example, "name": [u"%ubuntu%", u"%debian%"] to get all those items with either “ubuntu” or “debian” in their name).

Go get it here! (Note: Zeitgeist by itself doesn’t provide any user interface and shouldn’t be relevant to end users. However, a GTK+ interface depending on it should be available soon, followed by GNOME Shell and other products which will also use it.)

Related posts:

1. Zeitgeist since UDS
2. Zeitgeist is out!
3. Introduction to Zeitgeist 0.2’s API

2 comments
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: gnome, gsoc2009, Programari lliure, Ubuntu, zeitgeist

Installation

If you use Ubuntu you can grab packages from Zeitgeits’s PPA, else get the tarball and install it (# make; make install). Once installed, start “zeitgeist-daemon” and leave it running in the background (at first it may be busy for some time going through all of your Firefox history, but after that you shouldn’t notice it).

Zeitgeist comes with a public Python module providing some convenience functions. This module makes it very easy to play with Zeitgeist from an interactive Python instance, which is something I recommend you to do while reading this post. Thanks to it, you can get access to Zeitgeist’s D-Bus interface simply doing: from zeitgeist import dbusutils; iface = dbusutils.DBusInterface().

API Examples

For brevity, I’m only going to show some examples, but not explain what every parameter used in the call stands for. You’ll need to look at the DBus API documentation to completely understand them.

Also, I’m using Python for the examples, but the same works with very little modifications in any other language (as long as it has D-Bus bindings).

iface.GetItems([u'file:///home/rainct/todo'])
Returns a dictionary with basic information about the “todo” file in my home directory; uri, text (usually the filename), mimetype, tags, whether it’s bookmarked or not, etc. The values for some keys, like “timestamp”, are empty because they are only relevant for events, which are items put into a timeline (yes, this is a bit confusing, in future versions of Zeitgeist we will clearly differentiation between events and items).

iface.FindEvents(time.time() - 24*3600, time.time(), 5, False, 'event', [])
Returns the five last events (or less, if less than five events where logged within the last twenty-four hours). If you opened the same file twice, it will have two events; to avoid this change 'event' to 'item'.

iface.FindEvents(0, 0, 10, True, 'event', [])
Returns the ten first events ever registered with Zeitgeist which happened first (the sorting is always done by timestamp, not by insertion order).

iface.FindEvents(0, 0, 2, False, 'mostused', [])
Returns the last two events for to the items (URIs) which are used the most often.

iface.FindEvents(0, 0, 0, False, 'item', [{'tags': [u'zeitgeist'], 'bookmarked': True}])
Returns all events (when there are more than one with the same URI, only the most recent of them because of the 'item' parameter) which are bookmarked and have the tag “zeitgeist”.

iface.FindEvents(0, 0, 0, False, 'item', [{'tags': [u'zeitgeist']}, {'mimetypes': ['text/x-python'], 'bookmarked': True}])
Returns all events (when there are more than one with the same URI, only the most recent of them because of the 'item' parameter) which are bookmarked and have the tag “zeitgeist”.

iface.FindEvents(0, 0, 0, False, 'item', [{'tags': [u'zeitgeist', 'engine']}, {'mimetypes': ['text/x-python'], 'bookmarked': True}])
Returns all events (without duplicated URIs) which either have the tags “zeitgeist” and “engine” or, alternatively, are Python files (determined by their mimetype) and are bookmarked.

iface.CountEvents(0, 0, 'item', [{'tags': [u'zeitgeist', 'engine']}, {'mimetypes': ['text/x-python'], 'bookmarked': True}])
Same as above, but returning only the amount of results that the previous call would yield.

iface.FindApplications(time.time() - 3600*24*30, 0, [{'mimetypes': ['text/x-python']}])
Returns the path to the .desktop files of all applications which were used to manipulate Python files within the last 30 days (eg., in my case: [u'/usr/share/applications/geany.desktop', u'/usr/share/applications/gedit.desktop' ]).

iface.GetTags(0, 0, 5, u'a%')
Returns the five most used tags which start with “a” (case insensitive) in a list of tuples containing both the tag name and the amount of times it was used.

iface.GetLastInsertionDate(u'/usr/share/applications/gedit.desktop')
Returns the timestamp of the last event related to Gedit.

iface.connect('EventsChanged', callback_function); import gobject; gobject.MainLoop.run()
Calls callback_function every time a new event is inserted into the database or an existing one is modified. The callback function receives a list containing in the first place a string representing the type of change (“added”, “modified”, “deleted”) and in second place a list containing the affected events (or, in case of deletion, only a list with the deleted URIs).

Related posts:

1. Zeitgeist is out!
2. Here is Zeitgeist 0.2.1!
3. Zeitgeist Data-Source Registry

No comments
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: gnome, gsoc2009, Programari lliure, Ubuntu, zeitgeist

«««
With great joy and excitement, the Zeitgeist Team is proud to announce Zeitgeist 0.2 codename: okidoki), the first public release of the ultimate service for event logging (yes, there was no 0.1 announcement, we are too cool for that).

Zeitgeist is an “Event Logging Framework” that provides cross application awareness of the desktop’s activities.

You worked on a file, but you cannot remember where you saved it? You visited a web page about basketball three days ago, but you cannot find the URL in your browser’s history? No problem, this is where Zeitgeist enters the scene. It knows a lot about your activities and has a feature rich D-Bus API which allows GUI applications like gnome-zeitgeist, zeitgeistfs and others to present you your activities in a readable way.

Highlights of the 0.2 release:
* SQLite backend for central information storage.
* D-Bus API with support for storing and retrieving the data (with lots of filtering options!).
* Support for logging items and events from GtkRecentManager,
Firefox, Tomboy, Evolution and Twitter. Plus plugins for Epiphany and Bazaar!

All this has been possible thanks to (in alphabetical order):
Markus Korn, Mikkel Kamstrup, Natan Yellin, Seif Lotfy and Siegfried-Angel Gevatter!

Installing Zeitgeist

To install Zeitgeist just download the tarball, extract it, and then run:
$ make
# make install
If you are using Ubuntu it’s recommended that you install the packages from our PPA. Finally, you can start the daemon, like this:
$ zeitgeist-daemon
»»»

Related posts:

1. Here is Zeitgeist 0.2.1!
2. Introduction to Zeitgeist 0.2’s API
3. Zeitgeist since UDS

13 comments
© Siegfried-Angel Gevatter Pujals, 2009. | Permalink | License | Post tags: gnome, gsoc2009, Programari lliure, Ubuntu, zeitgeist