Computing related collection of stuff

I continued work on hg2mtn (which managed to import the whole onnv-gate repository just fine, stats will come later), and except for one issue, I’m feature complete: while I track renames now, I have to synthesize directory renames. This will work by looking for each rename if the source directory is getting removed and the target directory newly created. If so, assume that a rename from source dir to target dir takes place.

This might give some wrong decisions, eg. in this scenario: a/a is moved to c/a, while the other 100 files (that are sorted behing a/a) in a/ are moved to b/.  The proper rename would be from a/ to b/, but a/ to c/ is chosen. I could do majority "votes" and things like that, but they’d quickly get too complicated while still staying a heuristic.

The only issue that happens in that case is that a new file in a/ from some branch gets moved to c/ instead of b/ (where it likely belongs to) - but that’s easy to solve and still better than the current scenario where you’d have to move the file out of the deleted directory before merging to avoid a conflict.

After I implemented that one, I only have to clean up some more (I started already) to make a stable release, yay!

Darren Reed’s thoughts on the opensolaris community buildup - very interesting read!

It seems, that the "architects" came to similar conclusions already, given that the community/project stuff was reorganized a while ago. Unfortunately it’s still very rigid.

Of course, it doesn’t matter to Project Independence - we won’t ask for a slot on opensolaris.org, so we won’t have to mess around with their idea of "community".

Today I worked a lot on hg2mtn, a tool to import hg (mercurial) repositories into monotone.

The result of it is a first release that’s barely useful, but contains all the logic necessary to do it - what’s missing is userinterface, stability and general niceness.

After DIN (German national standardization body) voted in favor of OOXML, I looked for details on their website - They had an announcement, but not really in-depth information. What I found, however, was a contact form.

Given how some of those contact forms do not seem to reach a human (see my Qt inquiry), I didn’t hope for too much, but still spent some time writing down the questions I had. That was yesterday.

Today, just a bit more than 24 hours later, I got a response. Not too detailed, as some internal things shall stay internal, it seems, but definitely no templates, but some real human writing real answers.

I guess a proper summary of that answer is, that some of them wondered about the various issues that surround OOXML’s standardization effort (Fast track, two standards for one purpose, etc) and asked back at ISO. When it became clear that ISO will push this forward, they moved on with debating the standard according to procedures - what else should they do?

It seems that they expect more work (and probably changes) based on comments from the national bodies later-on in the process. I only hope that this expectation stays true!

The mail also stated that it’s possible for individuals to participate at the national level. I really have to look into this for future standardization efforts.

(Oh, and now that you know that they have real humans behind those contact forms, please fight the urge to spam them. Thanks)

So I’m back to Solaris again - given the massive growth of the software stack over the last few years, including several proprietary components (hello adobe flash), it seems getting harder and harder for smaller systems to keep up.

Even with all the troubles OpenSolaris has, starting from there still looks like an easier path than moving to a smaller system - sadly. Those were the days, where one developer could understand the whole system that he holds in his hands… *sigh*

Anyway, I promptly worked a bit on pmpkg, and implemented digests (as optimization, not as security check for now), updated some things here and there and I’m planning on lint functionality, both for package sources in pmpkg as well as the resulting package (necessary due to the dynamic design of pmpkg - we don’t require packages to match a given file packing list, but we want to enforce a certain set of constraints).

Also restarting is my effort to get into the ONNV tree - my toolchain bootstrap script is updated, and I think I’ll work on hg2mtn next.

So I installed dfly on my laptop now: 4 years old, keyboard broken, power plug socket somewhat broken, NIC socket somewhat broken, dvd drive broken (and removed by now).

How it’s done:

1. Download and burn the installer ISO
2. Boot on a different computer on the same network
3. Select the Network install server on the second screen (after language selection) in the Installer
4. Select the NIC to run it on
5. PXE-Boot the laptop
6. Get over the irritation of a NIC socket that has some defects (plugging in several times helped) - the dfly driver is more sensitive to that than the PXE driver that loaded the kernel
7. Install on the Laptop as if you worked from the CD (actually, you do)
8. Reboot Laptop
9. Remove CD and reboot the other system

Done

I have to say, I’m very pleased!

Ever wondered what "Linux" really is? Seems like its proponents aren’t so sure about it, either:

• When it comes to migrating people off other systems (eg. Windows), often it’s said that they should "use Linux" (no distribution specified, that’s an implementation detail after all, and Linux users are supposed to switch them all the time anyway)
• When it comes to security issues, bloat, and so on in those whole systems, some of the same people who encouraged use of "Linux" are quick to tell that a security issue in (say) Firefox is not a "Linux problem", as Firefox is not part of the kernel.

Conclusion: Those people who propose migration want their victims to use a kernel, without (unsecure) applications.

Yeah, right.. 

As part of this years Google Summer of Code I’m working for the LinuxBIOS project. My assignment is to make GRUB2 work as payload for LinuxBIOSv3 (the most recent version of LinuxBIOS). This means, that GRUB2 ends up in the flash ROM and gets loaded by LinuxBIOSv3 (which sets up the memory controller and things like that), so that GRUB2 can continue the boot process (provide a menu, load the system, etc)

One of the issues of that task is that GRUB2 so far only had a disk driver that uses the PCBIOS interfaces - highly compatible and somewhat nice to use, but on a LinuxBIOS system, these interfaces do not exist (and shouldn’t - being 20 years old and part of the 16bit interface).

The last few months I worked on getting GRUB2 to load at all, on using keyboard and vga text mode (which are also done via PCBIOS in GRUB2), and on providing a serial console (easier than I thought).

Today, I finally worked on the IDE driver - reusing code from OpenBIOS (somewhat related to LinuxBIOS), and managed to get it ported in 8 hours or so.

So, what’s left for GSoC for me?

• Booting a multiboot client doesn’t work for weird reasons. While this isn’t strictly part of my assignment, having a boot loader that can’t boot is kinda - pointless..
• Cleaning up the code, separating it out, manage all the licensing mess that the FSF generates for their projects and get the code in their repository (the changes I made on the LinuxBIOS side are already committed)
• The IDE driver I used provides an ATAPI interface already, but there are some small issues with accessing CDs (with block sized 2048 bytes) from GRUB2 (assuming all devices have blocks sizes 512 bytes).

flac: If the build of flac 1.2.0 fails, check if it is already installed in an older version and at least remove its headers!

sbcl: I enabled threads now, which survived basic testing. So finally a native threaded, compiled, secure, highlevel dynamic language on Solaris!

Trolltech decided to give more Open Source projects the possibility to use their fine Qt library - Unfortunately they seem to have forgot about the CDDL (but the MPL is in there).
I asked them for their reasons, probably they just didn’t realize that the CDDL would be a fine addition. Let’s see when/if they answer…