Stuart Herbert | Gentoo

I have to be up for work in six hours, and just haven’t had enough time to finish my PHP 5 status email for the gentoo-dev mailing list.

It’s useful to write status reports, as it gives you something in black and white to stare at and digest. For example, the PHP5 ebuilds don’t install the PEAR installer; we moved it out to the PEAR-PEAR package. But, when you look at it, the PEAR installer requires dev-php/php anyway, and it also requires dev-php/php to include libxml2 support, so I can see why a lot of PHP users don’t support our decision to move the PEAR installer out of the PHP5 ebuilds.

There’s a couple of reasons why. The first one is that the PEAR installer can be released independently of PHP releases. Having the PEAR installer as a separate package means we can get the latest fixes to users as quickly as possible. The second reason is that it’s against Gentoo policy for two or more packages to overwrite the same files. As policies go, it’s one of the ones that we devs can break and (so far) get away with, but I’m sure that, sooner or later, that’ll have to change. And we need to be ready for it.

Sebastian is working away atm, but when he gets back I hope to have a chat with him about how we want to handle the PEAR side of things - it’s much more his area than mine - and make sure the PHP 5 Status email covers what he wants. Until then, it’s best for me to hold back on the email.

After six months of treatment on a back injury, it’s really nice to be back at the keyboard and able to put more time into Gentoo once more. I wouldn’t be coming back at all if it wasn’t for the help and support I’ve had from many Gentoo developers. In particular, Kloeri, Tigger, Beu and Ka0ttic did a fantastic job of picking up things I wasn’t there to do. I can’t thank you guys enough.

So what am I back to do?

• Apache stabilisation: The apache team has done a great job in putting together much improved packages (and, for the first time, documentation!) to replace our previous attempts. The day is fast approaching when we push these packages through to stable. With my PHP maintainer’s hat on, I’ll be there to help out with reported PHP issues, and to get any necessary fixes back into the tree as quickly as possible. I’ve setup a repeatable test environment (based on VMWare Workstation 5) which should help us minimise problems on the day.
• PHP 5: I’m the primary maintainer of our PHP 5 packages, and with me not really around, there’s still a bit of work to do to get PHP 5 stable. The problems aren’t with PHP 5 itself - many people are finding it to be fine for production use. But the PHP packages don’t sit in the Portage tree on their own. There are many packages which depend on it, and we have to make sure that the necessary is done to them before we can unmask PHP 5. We also need to make changes (and write documentation) to support a mixed PHP4/PHP5 environment. I’m about to start a discussion about this on the gentoo-dev mailing list.
• Recruitment for webapps: The webapps herd already has more packages in Gentoo than we can keep pace with. And, if you do a search on Bugzilla for bugs assigned to webapps-unmaintained@gentoo.org, you’ll see that there’s plenty more waiting for us. As a team, we need to recruit maybe up to ten more developers as quickly as we safely can. If you’re interested, the way to start is to catch me or rl03 on Gentoo’s IRC channels, and point us at ebuilds or patches you’ve submitted to bugzilla.
• Replacing webapp-config: webapp-config has been very useful. It’s proved that we can add, upgrade, and remove web-based applications via a package management tool. It’s allowed us to gain a lot of experience in what works, what doesn’t, and why. Bugs aside, it’s proven to be a step up from trying to tackle the unique requirements of web-based applications through Portage alone. But it’s also shown us that some of my choices were the wrong ones, and fixing them will take a little more than speeding things up by switching from bash to Python. Once PHP 5 has been stabilised, I’ll return to working on a replacement, which we’re expecting to call the vhost-tools.
• NX Server 1.5: For the relatively few of us who use it, NoMachine’s nxserver is very useful indeed. They’re currently testing the next version which, for the first time, will allow suspended sessions to continue to run - which for the first time brings it on a par with Windows Terminal Services. That’ll be nice … I’ll finally be able to leave my desktop running here, and just remote desktop in via nxserver to check my email. Heck - I’ll be able to join in discussions on the gentoo mailing lists during the day

And, when that lot’s done, if no-one else has picked it up, I’d like to get Xen into the tree.

With the failed hard drive replaced, the machine has gone from being unbootable back to crashing regularly. The next step is to strip out all non-essential hardware, and to switch the machine over from the PCI NIC to the onboard NIC. Isolate to eliminate.

Mmm … let’s see if my blog is still up and working on Friday afternoon

Looks like I spoke too soon Monday night. I get home Tuesday (after watching Batman Begins … enjoyed that one) to find the machine down again.

Turns out it wasn’t the motherboard (although I’m really glad to be rid of that SiS chipset) - it was the primary drive. Not what you want to happen to your web and imap server.

Fortunately, when I built the box, I’d put the disks into a RAID1 setup, and the second disk hadn’t died (although I’m taking no chances, as both disks were of the same make!) Unfortunately, I hadn’t put the boot partition onto RAID1 - but with a minimal install CD to hand, it didn’t take long to get /boot recreated.

Most of the time was taken reading up on how to recover from a physical disk failure, and then copying all the data from the surviving RAID1 disk onto one of the two new disks I bought at lunchtime. (There’s no way I’m risking that data by trying to re-use the surviving disk. Much safer to make a complete copy onto a new disk, and then rebuild the RAID1 array by adding a second new disk. Anything goes wrong, all I’ve lost is time - not the data).

I haven’t actually rebuilt the RAID array yet. Gareth recently gave me a demo of EVMS, and I figure this is the perfect opportunity to convert this box over to EVMS. Before I do that, I need to do a bit of reading and satisfy myself that recovering data from an EVMS-managed RAID is every bit as easy as recovering data from a normal Linux RAID.

Dead motherboards - grrr. Had the one in my mailhub/webserver die last week. Swapped it for a nice shiny Asus (which my Dad didn’t want any more), and everything worked first time

I imagine I’ve quite a backlog of email. There’s very little chance of me reading any of it, so if you sent me something important, wait a couple of days and then send it again.

For the curious, packages for php-5.1.0-beta1 are now in the tree. It’s too early to have ironed out all the compilation problems. If you come across one, please file a bug (especially if you have a fix).

They install on my dev box just fine. Once they’re on the rsync mirrors, I’ll see how well they fair on my new test box.

Assuming they work fine, I can get the php-config package done, and then we can start looking at PECL packages for PHP 5.