Mohammed Sameer's Online Aggregator

Last year, after many years of peacefully living the whole year in GMT-3, our government decided that Argentina should use DST again. This was done in a rush, but patches were written and applied everywhere to have a correct timezone.

Fixing the problem was not enough, tzdata's upstream decided to predict the future:

# From Paul Eggert (2007-12-22):
# For dates after mid-2008, the following rules are my guesses and
# are quite possibly wrong, but are more likely than no DST at all.
Rule    Arg 2007    only    -   Dec 30  0:00    1:00    S
Rule    Arg 2008    max -   Mar Sun>=15 0:00    0   -
Rule    Arg 2008    max -   Oct Sun>=1  0:00    1:00    S

Well, back in December 2007, October 2008 might have seemed like a long time in the future, and they assumed that sometime in the middle the correct date would be announced... But, predictably (knowing Argentina's current government), no DST has yet been announced.

However, Paul Eggert's guesses had propagated to almost all UNIX distributions, so that at 0:00 Sunday Oct 5th, all our computers were suddenly one hour ahead of time.

Affected systems: Debian, Ubuntu, Gentoo, Red Hat, OpenBSD, FreeBSD, Mac OS. Devices: ipod, maemo, Nokia N95. Websites: Clarin (a major newspaper - still broken 36 hours later), Gmail. And probably many others that I don't know of.

I even read that Windows was affected, although I don't know how that might be possible, since as far as I know they don't use tzdata.

Anyway, not satisfied with how they mistakenly predicted the future before, a few weeks ago (not enough time in advance to fix almost any systems, only sid had this release), they added this change:

# From Alexander Krivenyshev (2008-09-05):
# As per message from Carlos Alberto Fonseca Arauz (Nicaragua),
# Argentina will start DST on Sunday October 19, 2008.
#
# http://www.worldtimezone.com/dst_news/dst_news_argentina03.html
# http://www.impulsobaires.com.ar/nota.php?id=57832 (in spanish)

Rule    Arg 2007    only    -   Dec 30  0:00    1:00    S
Rule    Arg 2008    max -   Mar Sun>=15 0:00    0   -
Rule    Arg 2008    max -   Oct Sun>=15 0:00    1:00    S

So, they decided to take the word of a guy from Nicaragua (no bad feelings against Nicaragua, but I think that this kind of stuff should be informed by people from the affected country), and from a couple of articles that say that "According to some sources, we might have to change our clocks on the third Sunday of October". Would you change tzdata sources with such information? I definitely wouldn't.

Anyway, Aurelien Jarno has already uploaded fixed packages, with no assumption regarding when the DST will happen, which is the sensible thing to do in a case like this. Thanks Aurel!

Finally, after more than a year of preparation, and six months of very very hard work, DebConf8 has come and gone. Even if I'm not yet completely recovered from all that stress, I'm good enough to feel really happy about how things turned out.

DebConf8 was a great success. We had great talks, many opportunities for developing interesting ideas, a lot of social interaction, an awesome video team that allowed more than 200 people from all around the world to be part of the conference even if they weren't in Argentina, and in general almost everyone had a very good time. It was really nice to have so many people from Debian over here, and it was specially nice to see them working and enjoying themselves so much.

This was all possible thanks to our sponsors, thanks to the many hours spent during the previous months both by the DebConf orga-team (the usual suspects) and specially by the local team, which includes Tincho, Damián, Romanella, Maxy, Sebas, Zero, Mendieta, Dererk, Melisa, Angasule, Lisandro, Nueces, and also thanks to the all help of the volunteers that came to work with us during DebCamp and DebConf, which include Tomás, Tinchito, Mónica, Lucas, Germán, Diego, Fefu, Nicolás, Martín, Marcos, Hernán, Alejandro, Matías, Rodrigo, Alberto and Joaquín, and finally, DebConf wouldn't have been the great event it was without all the people that managed to travel thousands of kilometers to get here. To all of them, thanks, for making DebConf8 such a great conference

Now, at last, DebConf8 is over (although there is some stuff that we still need to do before we can really forget all about it), and life goes on. Today, I did my first NMU after a long time. I'm particularly glad to have time for fixing bugs again, but I won't lie, I'm also extremely satisfied with how DebConf8 turned out.

See you in Extremadura!

Caution: Latin American rant ahead.

With my Latin American Spanish keyboard (xkb code "la") I can type in:

• English, Italian, Dutch, Basque: abcdefghijklmnopqrstuvwxyz
• Spanish, Galician, Quechua: áéíóú ñ ü
• Catalan/Valencian and Occitan: l·l ç ïü àèìòù áéíóú
• French: àèù âêîôû ëïÿü ç
• German: ß äöü
• Swedish: å äöü é ç
• Danish and Norwegian: æ ø å
• Icelandic: áéíóúý ð þ æ ö

But not in:

• Portuguese: I can't type ã or õ (no dead_tilde)
• Guaraní: I can't type ãẽĩõũỹg̃ (no dead_tilde).

It turns out that for some people it's more important to have 3 (yes *THREE*) asciitilde (~) in the Latin American Spanish keyboard, than to allow people to write in the language spoken by 51% of South American people, or the second official language in Paraguay (spoken by 94% of the population).

The same thing happens with the Traditional Spanish keyboard (code "es"), which was initially thought only for Spain, but is now widely sold all over Latin America. It includes 2 asciitilde, but no dead_tilde.

I think this is outrageous and I'm very very pissed about this. As can be seen in the posts I've made to the bugs in Debian and in FreeDesktop.

However, it looks like we Latin Americans are overwhelmed by the amount of Spanish people in Free Software (particularly in Debian) who don't care that Brazil is the biggest economy in the region nor that other native american languages can't be written without a dead_tilde.

For the record, there are some other European languages, that can't be written with the Latin American keyboard, such as:

• Finnish: I can't type žš, only åäö (no dead_macron)
• Polish: I can't type ąę ż, only ł ńśź
• Czech: I can't type čď ěňřšťž, only áéíóúý
• Romanian: I can't type , only âî

But in this case, it makes much more sense to not be able to write those than not being able to write Portuguese or Guaraní, and it's not like there are 3 macrons and no dead_macrons, there are no macrons at all (same for all others).

After reading Christian's post about the new ISO 639-3, I thought about the "what is the country in the world which as the highest number of languages listed in ISO 639-3" question, and thought, "It must be India or China", and sure enough they both have a high language count (428 and 236 respectively).

However, after clicking around a while, I found out that Nigeria has 510 languages. I thought I had found the highest one for quite a while, until I got to Indonesia, which has 742 languages, and I thought "it's not fair, that's much more than just a country, it's a huge group of islands". Not having learned my lesson, I was quite astonished when I finally found out that the country with the highest number of languages is Papua New Guinea, a neighboring group of islands (although not so big as Indonesia), with 830 languages!

In Europe, the coutry with the highest number is Turkey (36), followed by Italy (33) and France (32). It looks like Europe has gone a long way after that stupid Babel incident :).

In America, I thought Argentina would be quite up in the list, since we do have a lot of native groups, but it turns out we only have 27 languages and we are on the 9th place. The country with the most languages is Mexico (298), then USA (238), then Brazil (235).

And, after all that clicking around the site, I found a very interesting map, that has one red dot per language in the primary location of each living language. It's quite amazing to see the big red New Guinea island. I wonder what happened there that led to the creation of that many languages.

In a sudden rush of stupidity, the Argentinian government decided that we should change our timezone to include DST.

For those that don't know, Argentina lies almost completely in the GMT-4 zone. 20 years ago we used to have DST, switching between GMT-4 and GMT-3. But since 1990 we've been using GMT-3 as the permanent timezone for our country. Thus, noon happens at 13:00 (or even later in more western parts of the country).

Now, since we are in energetic crisis, our government decided that we should go back to DST, but instead of GMT-4 and GMT-3, we are going to be GMT-3 and GMT-2. This means that during the summer noon will happen at 14:00 or later.

Not only this, but they decided to do this on December 21st, passed the law on December 26th, and published it on December 28th. And the day of the switch is December 30th!.

Thanks to the quick work of Clint Adams, a patch was provided, applied and uploaded to unstable today. It's currently in incoming, but it'll be in unstable in less than an hour. After the package was uploaded, I patched Etch's version so that we could upgrade all our servers.

So, in case you need to take Argentina's stupidity into account, you can currently download tzdata_2007j-3_all.deb from incoming (soon enough directly from unstable). Or download tzdata_2007j-1etch2 from:

deb http://volatile.debian.org/debian-volatile etch-proposed-updates/volatile main

Ever since I stopped doing webpages, I've been missing having my own Rhino (the one I used in the past was not mine), but since I wasn't doing any JavaScript, it didn't make sense to spend U$S 50 on a book I wouldn't use.

Now, with the interactive web inteface for Debbugs being accepted as a Google Summer of Code project, I finally decided to buy one for me. I ordered it yesterday from Amazon.

I hope it gets here soon.

So, last week was a bit bumpy since I was sick for a couple of days, and then had a power outage on Friday and Saturday, thus not being able to keep up with my daily RC bug fixing, but I've been catching up since then.

I've uploaded patches done by Arjan Oosting, for packages that wouldn't compile with autoconf2.6: #379812 (kde-style-polyester), #379813 (kxmleditor) and #379815 (klog).

I uploaded a patch by Martín Ferrari for lilo: #374477 (use MAKEDEV instead of mknod at postinst), and another patch by Martín for courier-authlib #378571 (fixing the permissions of /var/run/courier/authdaemon).

I uploaded a patch by Andreas Jochens for ntlmaps: #379700 (fixing the build-dependency). A patch by Luca Bruno for predict: #379495 (fixing a change of location of forms.h). And a patch by Mike O'Connor for stardict: #379152 (fixing a misuse of size_t that made it fail in 64bit architectures).
Stardict had a problem with rpath that made me get to know the LIBTOOL_IS_A_FOOL hack, and I had to apply that, so that the binaries were not screwed up.

I also uploaded a fixed version of yacas, that included patches from Arjan Oosting (fixing #379261 and #379895) and Braun Gabor (fixing non-RC #295413).

However, it's not like the only thing I've done is upload patches done by others. I've also done some patches myself:

• #379275, fixing fillets-ng's FTBFS.
• #379486, fixing a bashism in cbuild.
• #379200, changing PWD for CURDIR so that sudo works for building bluez-utils.
• #377781, fixing monotone so that it builds with the latest gcc, although it took a new upstream release for it to build in hppa.
• #378173, adding some checks for NULL, so that the new libsdl1.2-image does not make programs segfault
• #381248, dropping PAGE_SIZE in favour of sysconf(_SC_PAGESIZE) in libsdl1.2 so that it builds in powerpc

The amount of bugs currently affecting the next release is 247. It's not a secret that we are a bit behind the schedule (we should be at less than 200 bugs by now). But it's also totally possible to get back on track if we work together on fixing the current bugs and we stop uploading unneeded new releases that trigger new transitions.

Following Sesse's example, I decided to start bugfixing one RC bug per day, so, this is the result of my first three days at bugfixing:

• #377652, by fixing the source code to look for rgb.txt in the correct location.
• #377365, by finding out that the bug was not present anymore.
• #377608, by updating to the new usage of libavformat-dev. Actually, both Sesse and I fixed the same RC bug at the same time. :-\

It's quite rewarding to think that if you fix 1 bug a day, from today till November, 30th, you can fix about 140 bugs. With a couple of us doing this, we could tackle not only the RC-bugs but also the important ones (but we have to coordinate, so that we don't overlap again).

Following the trend, these are the bugs I squashed yesterday and today

• #377694 (graphviz), by backporting a lua test for the configure script.
• #377990 (libopengl-perl), by adding a build-dependency for libxi-dev.
• #377991 (freecraft), by applying a patch by Adriaan Peeters.

As the other day, the last bug was actually fixed twice (Bas Zoetekouw also prepared the NMU). The main problem here is that the BTS is being slow, and it's difficult to know that someone else has already fixed a bug.
So I guess that if you are going to join the the "fix one RC bug per day" movement, you'd better join #debian-bugs on oftc, and say which bug you are working on, so that we don't overlap that much.

On my last post, I stated that there were a number of RC bugs affecting etch, which was not completely accurate: I was counting only those affecting both etch and sid, this is to say those packages that are in etch and that still haven't been fixed on sid.

It was pointed out to me that a number of packages (around 70 today) are waiting for perl to go into testing, so that they can go in too, and many of those fix RC bugs. In the end, a lot of RC bugs were not being marked as fixed in etch because of perl failing to build in hppa, mips and mipsel.

After learning this, I decided that this perl bug needed to be fixed: I gathered a small team, including my guru friends, Damián Viano and Martín Ferrari, and between the three of us, we worked on fixing this nasty bug for three days.

After a lot of time spent compiling an recompiling perl in an hppa box, we finally found out that the problem was due to a particular optimization flag (-fdelayed-branch) in gcc-4.1. Compiling the conflictive source file without that flag makes the bug go away.

So, we submitted a patch and are now waiting happily for Perl's maintainer Brendan O'Dea to make the fixed upload. Hopefully, the number of RC bugs concerning etch will go down quite a lot during the next week, as the affected packages are able to migrate.