On a somewhat unrelated topic, I would like to remind the readers of this wlog that next weekend, winter time will enact, i.e. you’ve got to add only one instead of two hours to the current UTC time to be able to talk to your “nōn-CS” neighbours. I’m adding it here since I know I would forget it otherwise, and I just saw it in a newspaper I read at the imbiss.
Well, I got closer. So close that dual-boot CDs are possible with the new system. So close that I got everything in place to make both baselive and dualive CDs. But makefs(8) let me down and threw an assertion on – how gdb(1) helped me to find out – a deep directory, usr/libdata/perl5/i386-mirbsd/5.8.8/auto/B/C, thinking it adds the same directory a second time to its node tree (aborting thusly to not end up in an endless loop later). However, adding the allow-deep-trees mount option did not help. So I’m stuck.
There are some alternatives. The one which I like the most would be to further change the Sun disklabel to not only look like an i386 MBR to the BIOS, but also contain a “partition table” with only one partition of type 0x27 (MirBSD), starting at the chain sector (24, in our case) and being two sectors in size. The second one would then contain an i386 disklabel. A 4.2BSD FFS filesystem (created with makefs(8) which should work better there) would just be added after the ISO 9660 filesystem containing the boot stuff (and possibly, the /v10 directory, i.e. nothing less and nothing more than the contents of a normal dual-arch install-only CD, plus the Live kernel, which is just a GENERIC with root set to cd0f… since we’ll be using UFS then, we don’t even need /dev on a ramdisk any more). However, I do not exactly know how this would behave if we mix a filesystem using 2048-byte sectors with a filesystem using 512-byte sectors on the same medium and expect it to work both when burnt on a CD using 2048-byte sectors for the entire medium and when put on a CF card / USB stick / HDD / etc. using 512-byte sectors for all of the two filesystems. The Sun disklabel would also have to be adapted, unless we want to hide the i386-live part from it (or are too lazy to show it… but we’re perfectionists, sometimes).
I know from others that they put /usr on an FFS inside a vnd(4) file instead, but this has not only heavy performance issues, I also know our vnd(4) to behave slightly buggy on media with sector sizes other than 512, with OpenBSD having fixed some of that. I would thusly like to avoid it.
Just putting FFS on the CD is not an option either. Oh, there are endianness issues as well, so there’s probably no point in having the i386 FFS slice being accessible by sparc, since people in all of the BSD camps still haven’t understood how uncool they are (our rewritten elf2aout(1) being a prime example that it is not only possible but also highly useful to have such tools do internal endianness conversions as needed).
Now I either need an OpenBSD guru telling me how to accomplish it, or have to experiment. *sigh*
I’ve built i386 and sparc intermediate snapshots. We’ve been fixing things lately, so that another combined DuaLive CD is not so far off, before the more weird things (such as the 4.4OpenBSD merge) start to happen. On an unrelated side note, my SPARCstation 20 is crunching for BOINC at the moment, at about 75 MIPS only though…
Maybe LLVM 2.4 and llvm-gcc 4.2 will hit the MirPorts Framework now, because we need a recent compiler for some things (Objective-C, or to compile Webkit), and it’s better in terms of licencing (BSD + GPLv2), quality and portability than FSF GCC. Maybe someone ports Iced Tea so that comrad will be happy. I don’t know. It’s all our spare time.
There are a few things in libpng that need to be taken care of first though, but the other recent security issues (ftpd(8) must die anyway, IPv6 ndp, …) were handled. Maybe others, let’s see.
Ah, and of course I need to get the Live CD functionality fixed and switched to use NetBSD® makefs(8) instead of J�rg’s mkisofs first.
I still don’t believe in Web 2.0, Communities, etc. For example, on Ohloh (I still need to take care to not write oh lol!) there are quite some projects, but “nobody” (well, almost) uses them. CacheWolf, the geocaching software of choice, starts a conversion from the Ewe VM to the supposedly superiour Eve VM, but I yet fail to see success, still people want to already switch over the svn trunk, and only bugfix in the 1.0 branch… as if anything were usable before next spring, and a full replacement before end of next year or so, guesstimated………
Jonathan complains about the brokenness of software, libobjfw is suffering from GNU libobjc runtime bugs (a layer deeper, that is). I so feel with him, I know the feeling so well… sometimes it would really be better to be a cat. I also quite met with Azraël, Florian’s (my best friend) animal half – with approximately 13 kg much more of a beast than the neighbour’s cat I nicknamed “Mir” (just to name it after a BSD) and behaving much more “cool”, but very nice.
Hi everyone, mksh’s current development version should be tested on as many platforms as possible before a new release. This mostly means:
$ CVS_RSH=ssh; export CVS_RSH $ cvs -d :ext:email@example.com:/cvs co -PA mksh $ cd mksh $ (sh Build.sh -r && ./test.sh -v) 2>&1 | tee build.log
Afterwards, please mail me (tg@) the build.log file, play around a little with mksh(1), see if it works. You can use a relative path to the source directory for repeated builds and also pass different compilers or flags as shown below, but please remember that passing CFLAGS or LIBS overrides the defaults, which you thusly should include:
- CC=gcc ksh93 ../mksh/Build.sh -r && ./test.sh -v
- CFLAGS="-q64 -O3 -qstrict" bash ../mksh/Build.sh -r && ./test.sh -v
- CFLAGS="-xO2 -m64" /usr/xpg4/bin/sh Build.sh -r && ./test.sh -v
You are, of course, invited to join the mailing list discussions regarding features, bugs, and behaviour, or even send in patches for missing items or architectures. I’d especially be happy if someone made Minix 3 work… pdksh runs there, mksh didn’t last time I tried (with gcc, but once that’s up ACK should be easy). Or Plan 9.
If you’re on Ohloh, please add mksh to one of your stacks.
This may come shocking to you, but, yes, I indeed agree with Richard M. “FSF” Stallman. While I respect him quite a lot, similar to Bill Gates, I usually do not. But this time, there’s something which makes me, wholeheartedly:
At Chaostreff on Tuesday I learned that RMS protests against cloud computing (via OSnews). I actually had to learn this new term first, but after I did, I can only join the fun, despite so many voices against it. It’s really bad having few people control a lot of users, and everybody using it without a second thought¹². The Terms of Service³ on these “Web 2.0” thingies, as I called them for myself until now, are usually quite problematic as well. You should not trust everybody, even without being paranoid.
① Yeah, I know I’m still using Google’s search engine and Maps service. But that’s a voluntary choice, plus there is no (usable) alternative, so I’ve got no way around it due to my requirements. However, I do not have a Google account.
② Interestingly enough, Benny said on the phone that he agrees with me on this one, despite using a GMail account. Pardon, for my German readers (this wlog entry was drafted in Switzerland), that’s Google Mail, just to make this clear.
③ Ever tried to take back information, cancel an account, get rid of their spam, read through the EULA, get someone to change things which suck, get some knowledgeable support person, etc.?
Of course, RMS’ arguments are also all valid. While I do not like the “sort of communistic” licencing approach of the FSF, and how they use the GNU FDL to restrict people while claiming that really all information should be free (I wouldn’t want that either, but that’s mostly for personal reasons or privacy), I’m positive that we should think twice about the code we use and the data we give.
I still am amazed at why people use so many kludgy web browser applications in favour of local applications. We should probably improve the situation regarding local backups… BoxBackup has an interesting approach, encrypting them. And, gosh, someone wrote a cloud shell… using $favourite_name (SOAP, XML, XMLRPC, AJAX, insert whatever disgusting technology is used at the moment) instead of Unix IPC via Pipes to connect “applications”. Gah! *shudder*
Today, I hacked some more on mksh’s command line editing modi. They were probably written independently, so Benny’s request for vi mode cursor keys is unfulfillable reasonablily. I’m not quite done, there are some more patches to create, including one to the mksh(1) manual page, but that will come. It’s a hard job, and on a headache-clouded november day like this (no pun, look at the weather!) getting a hacking mood, as well as anything else done, is difficult. I did manage to catch a few sun rays, but… not much. Besides, sitting in front of a monitor all day makes you wish you had learned something else, even if everthing else goes smoothly – which it usually doesn’t…
I tried to make pax(1) compile without the -DLONG_OFF_T define… turns out that all the quad_t types and defines are not existant on Debian. Autoconfiguration and #ifdefs suck if you can do it without, but this will take some more time.
I had an interesting discussion about light bulbs with the MidnightBSD
people yesternight. Mine mostly changed to power saving ones I got for
free for switching to online invoices from our local provider. Neat. I
do keep traditional bulbs in the places where light is needed only for
very short times, like the entrance area, and frequently at that.
We often have interesting topics to discuss, e.g. languages, chocolate, things that suck… not only computing things, mind you ;)
I just use …/wlog.htm as my “blog” URI now if someone asks, even though this is not a blog. Symlinks rock.
While here, I’d like to invite you to try out mksh-current from CVS, as several (minor) bugs have been fixed. There are still some new features pending for another release.
Zaterdag 13 September, Baarn, Nederland. Iedereen wacht op me. Spijtig dat ik niet kan komen. Ik wilde met een vriend die een auto heeft naar Baarn gaan, maar die komt op 10:30 ipv 09:00 CEST, en dan ben ik ook nog niet klaar met wakker worden, koffie drinken, zaken samenzoeken, voor’t weekend inkopen, enzo. Op 12:00 zitten wij in zijn auto… maar dat zijn nog 2½ uren (of meer) tot dat wij aankomen kunnen, dus wordt besloten, er niet meer te gaan. SFD was vanaf 12:00 tot 17:00… voor 2 or 2½ uren moeten wij 5+ uren in’t auto zitten, en dan is nog niet zeker of wij het Brandpunt vinden. Slechte planing. In het duits is er een expressie voor dat: “Verpeilungsfaktor” – Wij zijn dan in’t zwembad geweest.
Groeten an iedereen die daar was… volgende keer dan. Echt jammer dat het niks wordt. Ik geloof ik had te weinig geslapen verleden week, door mijn nieuwe baan in zwitserland.
Wij zatten dan gisteren nog in mijn favoriet kroaats restaurant, er wordt besloten dat ik de volgende tijd minder aan mirbsd zelfs werk, maar het nieuwe server klaar maak en de VM zo installeer dat ze een takeover van de oude www.mirbsd.org vhost kan maken. Ik doe mss eerst nog een snapshot, maar die komt dan zonder Live CD (die moet ik opnieuw maken doordat wij geen mkisofs met al zijn fouten meer gebruiken). Benny doet in de tussentijd meer met ports, bv. fixes voor ccache en distcc (zodat mijn SPARCstation cluster packages kan bouwen ☺) en enige Qt4 ports. Nadat de twee server’s klaar zijn kunnen wij dan verder met mksh, mirmake en zo doen, wanneer het werk nog arbeidsvermogen overig laat (da’s soms moeilijk). gecko2@ doet goeie werk met het opzetten van VMware voor eurynome, ter gebruik stellen van een IPv4, enzo.
Wir haben jetzt ein Blumenkohl-Icon fürs iPhone, Dank an die RUBIs.
Google ist definitiv evil. Ich vermeide sie auch so wie es eben geht, aber bei zwei Sachen habe ich keine Alternative, da sie entweder einfach nicht existieren oder saugen: Suchmaschine (ja ich hab sie alle durch) und Karten (bin halt Geocacher).
Während ich die 2⁶ Caches locker voll habe, siehe die diversen…
…Statistiken, fängt Benny erst an, immerhin 2³ hat er schon:
Kimnotyze macht nicht wirklich weiter…
… gecko2 kommt selten zum Cachen, aber hatte letztens sein Vélo mit,
sodaß er sich mit Kabelaffe und mir auf Tour begeben konnte:
Und der Dr. Pfeffer findet in Bremen nichts mehr… wär er bloß mal
nicht umgezogen, jaja… gibt halt nur in Bonn schön viele Dinger:
Aber Benny hat echt fieses Anfängerglück, und ich bin fachblind.
Bruscetta sind superlecker, aber füllen, zusammen mit der warmen Mahlzeit, den Bauch sehr an… und wenn man dann vorher noch einen Mojito hatte, hat man gar keine Lust zu hacken, ist ziemlich platt… erst bsiegert@ dann auch mir aufgefallen. Wir sitzen gerade bei N 47° 31.936′ E 007° 38.043′ im Chaostreff Bāsel… und sind ziemlich platt. Aber irgendwas kommt trotzdem bei rum.
Kommt wer mit nach Baarn und ggf. trifft sich mit uns $sonstwo?
It seems that everytime I update cairo, another annoying bug pops up. This one is actually a regression: the exact same bug had been fixed in cairo-0.9.3, and now they reintroduced it. It seems that their code is only ever tested on GNU/Linux with all the latest libraries. Solaris and BSD systems are especially likely to blow up.
I have submitted the following text as fd.o bug 17450. Sorry if this wlog is becoming just a dump for upstream bugreports.
"I am trying to compile cairo 1.6.4 on MirOS BSD (an OpenBSD variant). The fontconfig included in the base system does not have the FcFini() function. As I gathered from searching the mailing list archives, this function is optional. However, on systems without FcFini(), the freetype backend is completely disabled.
"From reading the source code, I do not think that FcFini is really essential, and any calls to it are protected by #ifdef HAVE_FCFINI anyway. Thus, I made a small patch for the configure script to make a missing FcFini non-fatal for the freetype font backend.
"What's ironic is that this bug had already existed as bug 3951. It was fixed in 0.9.3."
Six years ago, I started collecting patches against OpenBSD, a bunch of 4.4 elitist snobs since the 1980s (or so they say). Said patchkit should eventually become MirOS BSD and the MirPorts Framework, along with a couple of more or less well-known and successful subprojects. Time to celebrate (TGIF)!
IOW, I hacked a Debian package of mirmake, depending on latest mksh of course. It needs more upstream work to be useful and pretty, though.
Let me plug here an animated GIF found somewhere in the net, which
shows my opinion about Netscape/Mozilla/Fire…tapir quite well:
And, even more unrelated, MirOS is (of course) continuously improving. We now use Fedora Core 4 libraries for the linuxulator, by suggestion of the MidnightBSD people, which should improve some things (it already has opened the door for Opera-linux plugins such as AcroRead-linux).
(The best browser of all is still lynx(1) though.)
Hum. This year’s FrOSCon… sucks. At least the catering, which was so much better last year (remembering the Chili con Carne, as well as the Chili non (sin) Carne, fondly… these were produced by wbx@’s family members, which haven’t been included in this year’s planning so some other persons did the catering). And being waked up at 04:00 in the morning due to a call on the mobile phone, for taking care of some drunken booth slave *grml…*, didn’t help either.
The evening’s social event also sucked totally. Since it had been raining until shortly before, they decided to have it inside, except the barbecue of course, but didn’t adjust the volume of the music played to the environment, which caused me to leave the MirOS/XF86 booth in favour of the Debian/grml booth, talking to Mika and Joey… who, like me, didn’t quite like being LITERALLY punched into our stomach by the basses. I left relatively early then, especially as the cocktails (Vutral brought me one, since I couldn’t go near the counter due to the volume of the… whatever they call music) tasted pretty bad and had a (too?) high percentage of alcohol. But talking to the various people, not just Mika and Joey, was good, even though just being at the FH (university) during the social event was really unbearable and physically endangering one’s health.
Other than that, we had quite a lot of fun at the conference, as
usual. I still think it has chances to close up to FOSDEM, but they
will have to make sure the catering does not get even worse. While
FOSDEM does not have any catering, FrOSCon 2006 and 2007
had good food, and regressions count as malus.
Once I found Marc Aurele La France at the train station (which was quite a task), things went well – he even fixed a bug in our xdm configuration at the conference. The bug was inherited from OpenBSD, as usual ☺
Marc also said he enjoyed himself.
I hacked an Asus EeePC… MirOS mostly works (no NIC though), and the graphics card runs at 640x480 VESA… with the new 915resolution port of today, it might do the 800x480, but I can’t test now.
I took the chance to discover a geocache in St. Augustin yesterday,
but, while Benny suggested we (him, me and gecko2) go caching in Bonn
today, they seem to prefer hacking on the laptops (considering the
weather, this is not the worst idea though), so I got time to write a
wlog entry (too). I still have some things on my TODO, like fixing the
ports with unfetchable distfiles, but hey.
Taking the day before and the day after the conference off is A Good Thing™.
Next one will be the Software Freedom Day in Baarn, Nederland – where not even Wim will go ☺ But I need a car… parents don’t help even if you ask them once a year, but maybe Jonathan from (near) Aachen will join me, he has a car.
My Thinkpad X40 will probably arrive today or tomorrow. Sadly, I didn’t invest any time in evaluating the products before… I learned that the ‘T’ series has 2.5" HDDs instead of the sucky 1.6" HDDs, and a much more solid lid chassis, at FrOSCon. My decision to buy an X40 was based solely on the observation that it was “in” at most OpenBSD developers some years ago (and thusly would most certainly work well with MirOS). Don’t do that then…
Once it’s there, I got to set it up and continue working. It’s bad odem broke so badly, especially as my current contractor (employer, except not quite so) would rather have me working full-time on the project, which is not entirely possible since I still have a life here. But upon setting it up, I might upgrade and/or fix some of the ports, since I’m at it anyway. Plus I get /home encryption.
Marc agreed to merging as much of our X11 changes (both these inherited from OpenBSD’s XF4 module and our own patches) into the XFree86® main tree, some ifdef’d, and helping us migrate to the new 4.7 or upcoming 4.8 release (or probably, 4.8.99.01, since I don’t think the merge will be there in time for 4.8 proper). He was a fun guy and well understanding our issues. The most funny part however was Thomas from Sourcemage (SMGL), who also maintains the mksh spell in their grimoire, considering to retain supporting XFree86® (and modular X.org, but not the buggy monolithic X.org) iff it’s still actively developed, which Marc assured me it is.
Benny agrees that the “Kaiser’s BIO-Kaffee” is decent, despite its pricing of only 3.99 €/£ (may be my high quality milk too, though).
Update: looks like I’ve got a package to fetch from the post station.
In my recent entry, I already talked a bit about pkg-config and its framework support. It turns out that there is another problem. The --libs parameter can be "split" into its parts by using --libs-only-L, --libs-only-l, and --libs-only-other. The LyX configure script, however, did not care about the latter, only the first two. Which brings me to the mail I wrote to the pkg-config mailing list:
"while compiling LyX/Mac the other day, I came across a problem in the way frameworks are handled. On Mac OS, frameworks are a different way to pack a library together with its headers. Framework-related options are put in --libs-only-other, while the configure script was only checking for --libs-only-l and --libs-only-L. While I agree that this is broken behavior, it brings me to my question:
"The way I see it, -Fpath (framework search path) is equivalent to -Lpath for libraries. Thus, -F should maybe be included in the output of "pkg-config --libs-only-L". The same thing goes for the "-framework name" linker option, which is more or less the same thing as "-lname", namely linking against a certain library/framework. Thus, the -framework option (with argument) should be added to the output of --libs-only-l.
In other news, FrOSCon has it second and last day today. I am currently in the train on the way. The party ("social event") yesterday was quite good, even though the weather was not as good as last year—it was cold and humid.