MidnightBSD Developer Journal

We've had our ipv6 tunnel down since Thursday morning.  Apparently, some users of sixxs have experienced problems with static tunnels from the chicago POP.  I'm hoping that it will be resolved soon.

Over the weekend, I got an opportunity to update some mports wtih fetching errors per the last AMD64 0.2 magus run.  The results are as follows:

• untested: 610 
• pass: 1526 
• fail: 191 
• skip: 63 
• warn: 97 
• ready: 0

I just purchased a Toshiba L305-S5955.  It wasn't my first choice, and I had neglected to do proper research on a few points. This laptop comes with an Intel Celeron 900 2.2Ghz single core cpu, 2GB of RAM,  160GB hdd, Realtek pciE NIC (10/100) and a Realtek USB wifi 802.11b/g card.  That's right, the wifi is on the USB bus inside the laptop.  

Obviously, since we don't have native support for this wifi card and it's usb there is no way to use it.  Our ndis wrapper does not support USB wifi devices.  

Interestingly, the ethernet adapter does not work either.  This is one of Realtek's newer parts and it is not supported by rl or re.  Due to the pciE nature, I'm going to work on getting this integrated into re(4).

The laptop is fully supported in Ubuntu Linux, so we're certainly trailing on hardware support for newer systems.

The sad thing is the realtek card works much better in windows or ubuntu than my previous linksys pccard with a broadcom chipset.  (yeah i really pick em) 

Aside from the hardware support issues, I've been pleasantly surprised by the build quality of this laptop considering it's low cost.  It's not at the same magnitude as my old T30 Thinkpad, but it's good.  

All dns ports have been checked for security updates and updated including BIND 9.5 and 9.6.  I've removed BIND 9.4 from mports as it's included with the base system (current) and it's the oldest version still supported.  It's also three versions back at this point.  

For the few people using MidnightBSD as a server platform instead of a desktop, there is a security update available for named(8).  A denial of service attack is possible against BIND using dynamic update packets.  It causes named to exit. 

• untested: 443 
• pass: 1801 
• fail: 125 
• internal: 2 
• skip: 41 
• warn: 74 
• ready: 0
•  
•  
•  

mport.create seems to hang on bouml-doc for no apparent reason.  No disk activity is in progress.  It just consumes cpu cycles for hours.  This port works with the old tools.  

I'm going to attempt to rerun the ports with internals.  That appears to be a node problem.  

I recently purchased new video cards for my system.  I wanted to get the new cards working in CURRENT with the latest version of X from mports.  After several lockups and other odd behavior, I found that I had to disable the second graphics card (crossfireX in windows) in the X configuration along with removing the keyboard and mouse configuration.  I was then able to start the X server with the radeonhd driver. Don't expect acceleration yet.   

If you've had trouble building db45, db46 or db47 on current, try updating your sources.  A fix was included in the world that corrects the problem.  There have also been recent updates to fix crashing bugs with SSL enabled mports on current.   

www/browser3 is now building on some systems.  It's been tested with current in a limited configuration so far.  Please report success/failure with the port so we can improve it.  (It's unbranded F* 3.0.11)

net-im/pidgin has been updated to 2.5.8 which fixes several security vulnerabilities.

net-im/pidgin-twitter has been updated to .84

net-im/libpurple has been updated to 2.5.8 as well.

mail/linux-thunderbird has been updated and fixes a security vulnerability.

multimedia/gstreamer was updated to 0.10.23.  This port hasn't been touched for two years, so it's a big update.  The plugins ports will be updated soon.  This does not include the update for the recently reported png vulnerability, but it does cover several others.

Here's the top blockers list from the last magus run (151)

graphics/libmng: 193

devel/gvfs: 190

lang/gnustep-base: 107

graphics/libGL: 107

graphics/dri: 75

print/teTeX-texmf: 27

x11-toolkits/libgnomeprintui: 14

lang/ruby18: 12

java/diablo-jdk16: 8

devel/libgtop: 6

graphics/graphviz: 5

textproc/docbook: 2

multimedia/xvid: 2

lang/ezm3: 2

databases/db46: 2

x11-toolkits/pangomm: 2

lang/ocaml: 2

misc/gnome-user-docs: 2

devel/gdb6: 1

x11-toolkits/gtkmm12: 1

graphics/libgnomecanvasmm: 1

multimedia/xmms: 1

textproc/redland: 1

mail/courier: 1

astro/libnova: 1

devel/hs-haddock: 1

mail/sendmail: 1

java/diablo-jdk15: 1

net/jags: 1

archivers/p5-IO-Compress-Bzip2: 1

audio/linux-libogg: 1

print/hs-hscolour: 1

multimedia/ffmpeg: 1

java/jakarta-commons-lang: 1

MidnightBSD current includes ZFS (version 6).  Over the weekend, I got some time to kick the tires.  A usb based disk was setup with a ZFS pool and some informal benchmarking was done.  We also created a new ZFS kernel config file to help people setting up a simple environment.  Tuning is necessary with ZFS, but it's a good head start.

Results were promising overall.  Since a formal methodology was not used, I won't post results.  One thing to note is that magus will not work with ZFS. 

MidnightBSD 0.3-CURRENT has undergone some changes lately.  We've removed PCC from the base system.  Details can be found in UPDATING; we decided that the system compiler must support C++ and Objective-C to some degree.   That leaves us two options right now, GCC and LLVM + Clang (eventually)  One possibly factor will be what system implements the new blocks extension to the C and Objective-C languages from Apple, Inc.  We're quite interested in there work with Grand Central and components that aided that development.  (see the snow leopard website and developer area at apple)

There's currently a few gotchas in current including problems building world in regards to the mport / libmport stuff.  We're working on it.  Also, we're toying with including libffi in the base system in order to provide a more complete Objective-C environment. 

We ran a magus run for the first time using the libmport and mport package creation tools under development.  There were about 700 ports that were not built either do to error or non built dependancies.  We consider this quite good for a first time out.  A few bugs have already been fixed with regard to this run and we've also fixed some ports that came up with problems.  During the same time, we ran magus on 0.2 amd64 to help us with a baseline.  Results were quite impressive considering the lack of attention we've had with ports in favor of other projects recently.  

Our plan is to get current back in shape for everyday folks, fix as many ports as possible and then provide a snap with mport packges for people to try out.  This isn't going to happen immediately, but it's our next big milestone before a 0.3 release can happen. 

I also want to point out the mport database format can change and there could be some quirks.  Chris is still actively working on it.  

MidnightBSD current now includes the netpgp application and library from NetBSD.  The software was based on the openpgpsdk; we have had a port of that for a few months.  However, this is a significantly enhanced version with several bug fixes and active development.  

We will of course track changes as this is an evolving piece of software.  Thanks to Alistair Crooks for this exciting addition to the BSD community.

Learn more 

MidnightBSD 0.3-CURRENT now includes GCC 4.2.1 as the system compiler.  I4B (ISDN for BSD) has been removed from the userland; previously it was removed from the kernel.  

Several changes have been made to sysinstall recently, and I'm working on switching over to libarchive.  In 0.3, bsdcpio (backed by libarchive) is exec'd for each install phase.  0.2 used pax as cpio and 0.1 used GNU cpio which was removed for security (and GPLv3) reasons.  The installer now tries to figure out keyboard mappings from the user's selection of a country at the beginning.  Tape install features were removed as they were quite dusty.

One may ask why I'm even bothering working on sysinstall at all.  While I have started work on a new installer, I realized that it won't be at the level I would expect for 0.3 necessarily.  Having a tuned sysinstall is a good backup and possibly the only option for the sparc64 architecture at this time.  Also, it makes sense to work more in the problem domain with an existing piece of software and then improve upon it. 

I'm planning on releasing a new snapshot soon with GCC 4.2.1 and the sysinstall improvements.  I'd also like to create a new live cd (or dvd?)  soon.  I'm considering using xfce 4.x and possibly gnustep (if i can get the port using the system compiler quickly) on it along with midori.  We'll see what transpires.  

Lastly, I'd like to mention my current thoughts on pcc.  I was excited about it at first, but at this point I don't think it makes sense for a default system compiler.  We've got C++ code in userland, and with our Objective-C ambitions, it doesn't make a lot of sense.  I haven't decided if I"m going to remove it or not from src yet.  It very well may be disconnected from the build for the release unless I get time to fix it/update it.  Both Dragonfly and FreeBSD have been eyeballing llvm + clang.  I've got reservations about that just as much as I do about newer versions of GCC with GPLv3 licensing.  Some members of the BSD community are strongly against the license and certainly it hurts business interests with the various projects.  FreeBSD has a lot of corporate users and developers.  On the other hand, we want to move forward with several GPLv3 licensed pieces of software in the installation for the graphical user interface components.  So, even if i make some grand point with the src tree, it's lost with the package as a whole anyway.  I don't really care all that much about it beyond the limitations it imposes on use of my operating system. I've been contemplating this issue (and others) lately as I'm thinking of writing a very public roadmap.  This project has been moving forward mostly from some mini roadmaps I've detailed in my blog and this blog in the past, ideas I've had for years and work done by new developers.  The flexibility is great, but many people don't know where we're headed and that means we can't gain momentum with more conservative users. I have to admit that I like a bit of randomness to software I write in my own time.  It does seem like a fresh contrast to work.  However, this is an operating system project and should have a higher level of planning.  (and yes we do discuss major changes amongst developers)

If you refer to the wiki planning document for 0.3, you can see that we're working on a feature release.  The critical points for this release are the new installer, libmport / mport package tools, and the crazy src merge I've been working on.  The good news is that libmport and mport tools are in "pre alpha" state and they are ready for selective testing now.  You can find them in the src tree, but remember that the database and other aspects could dramatically change.  A large part of the "merge" is complete.  That leaves the installer.

Here's a big picture thought process as of right now:

1. 0.3 will contain the changes mentioned above and install either KDE 3.5.10 or xfce or windowmaker + gnustep.  Many of our developers have fallen in love with xfce (not me), none of us are very happy with KDE 4 and we feel it's redundant to ship KDE as PC-BSD and DesktopBSD do that.  Not to mention it doesn't integrate well with GNUstep.  For this release it will be a choice.

2. Most likely we're doing a 0.4 release which will focus on syscons and refinements to the installer.  I'm sure some hardware and userland software changes will be committed.

3. We're talking about going to 1.0 after 0.4 among the developers.  Regardless which it is, we want to have a binary update system in place by then.  Also, usb flash images suitable for installs and live cd/dvd functionality.  The latter I'd like for 0.3 ideally.  It depends on time.

Finally, I'd like to mention Etoile.  They've made some progress with Etoile and we need to get the new software running on our system via mports.  I'm not sure the timing will work out for 1.0 but we will still consider shipping Etoile with the system.  At this point, I think a Redhat style dual desktop setup is in order (choice during install?) with xfce and some gnustepish environment. 

At the end of the day, this project is no longer just about what I wanted when I started it.  Users and developer input is important to consider with any open source project, particularly an operating system.  At present, I've heard a lot of people push for either a very light weight desktop or a very mac os x like desktop.

We've been doing some spring cleaning on the website.  Most of the pages validate as XHTML now.  The footer and many links have been fixed.  The documentation pages have been broken up and expanded a bit.  

Committers: if you see any pages that need to validate or need some TLC, feel free to fix them. 

• untested: 296 
• pass: 1942 
• fail: 68 
• skip: 26 
• warn: 151 
• ready: 0

I ran magus on my desktop yesterday. Here are the results:

MidnightBSD is now a group on facebook. 

Current is building again.  The problems with the openssh 5.2 import have been fixed.  mksh was updated the other day as well from CVS.