Roy's Blog

A Hacker's musings on Code | Tech | Life

Nothing major - the 4.0 branch gets a fix to send DECLINE messages correctly (basically we forgot to add the IP and ServerID to the message) and the ntp.conf file path is tunable by NTP_CONF=/usr/pkg/etc/ntp.conf in /etc/dhcpcd.enter-hook. 4.99.3 got released as well with more fixes, mainly for 4.0 features which were broken. I'm no longer hopeful about getting IPv6 support in by the end of the year as I don't see where I'm going to get the time from. Ah well, it's not that urgent yet.On a side note, the new NetBSD server is working out well. It's very stable, the network throughput is better for downloading (as it's my router at home) and this site is more snappy :)

Continue reading...

And it was almost entirely painless!Just like my quard core as an issue booting NetBSD MP (I need to have a PS2 keyboard plugged in OR the PS2 port disabled in the kernel), it seems that it doesn't like ACPI on my server. No matter - ACPI is not essential there. After that hurdle is passed the system booted. I guessed the wrong interface driver for the on board ethernet and my 6to4 tunnel isn't working - but everthing else appears to be.What IS interesting is that with all the services I need running, I have 300meg memory free and no swap usage whereas with Gentoo Linux the swapfile was being hit after a few hours of use. Obviously a few hours hasn't passed yet, but I'll monitor it. Hopefully less memory is used on the whole :)As NetBSD comes with things like postfix in the base OS setting them up was trivial. Also, postfix uses the dovecot sasl authenticator without needed in a recompile, which means I don't have to use postfix in pkgsrc OR that gawd awful program cyrus-sasl. I've also configured Heimdal Kerberos and will play with that, as that comes with the base OS also.

Continue reading...

Well, someone told me that the Google Android platform uses my DHCP client, dhcpcd. Well, now that the code has been opened up, they do indeed have dhcpcd-4.0.0-beta9 in their git repo.I must be doing something right if Gentoo, NetBSD and now Google use dhcpcd :DI dunno about you, but that gives me a warm fuzzy feeling 8)

Continue reading...

I've updated Drupal on this machine from 5.11 to 6.5The main reason I didn't before was because the spam module didn't work with 6.x. I've been using the captcha module recently instead on 5.x and noticed that it along with my other modules now worked on 6.x :)The upgrade wasn't entirely painless - image module needs to be upgraded to 5.x-2.x before the 6.x upgrade.Due to some faulty data, I lost my image gallery contents, but as I'm a real man and backup, restoring it and fixing the upgrade was fairly easy.Also, node_revisions lost the reference to the right sequence.I think everything now works as before, but if you spot something wrong please let me know ASAP 8)

Continue reading...

One of the nasty hacks left in OpenRC is that the init.sh script (ie, the sysinit runlevel) runs a lot of hardcoded stuff at boot, such as udev and mounting of various bits in /sys, /dev/pts and /dev/shm.Later today I'll be comitted a VERY big patch that allows the running of ALL these bits as seperate init scripts in a real sysinit runlevel. The only required change outside of OpenRC is a new init script for udev which I'll link here when comitted. Scripts for devfsd and mdev will also be required, but it will be up to others to make them.EDIT: This is now committed. You can grab a udev script and comment on Gentoo Bug #240984This is have the side effect of removing the concept of coldplugging and we're just left with hotplugging.Coldplugging can still be done, but it will be a pure udev configure option and not an OpenRC one.The patch also make a few changes internally to librc where we now return an empty list object instead of a NULL when nothing exists.This makes the code a lot smaller elsewhere, and easier to read. Interestingly we actually malloc slightly less through the running of rc than prior versions.Also, some cosmetic changes will be made as well with more to come later.The last thing to do is then look into a shutdown runlevel for handling the hardcoded power monitoring stuff, or an alternative solution.Then we can finally think about making this sucker stable!

Continue reading...