" /> preed's blah-blah-blahg: March 2006 Archives

« February 2006 | Main | April 2006 »

March 31, 2006

Descend and maintain three-thousand until established on the localizer...

it's really hard to context-switch from lofty, 40,000 feet discussions of organizational values, communication dynamics, innovation vs. maintenance considerations, requirements discussions for various development support systems, and how we'd like to fix what's broken and maximize what's not to... "respin this build" and "that Tinderbox is broken" and "Where are these builds?" and "That tag is set incorrectly..."

Talk about an "emergency decent" to meet a release schedule...

March 28, 2006

Drive-by

Mac universal binaries for Bon Echo Alpha 1 are finally available.

The source tarball too.

March 21, 2006

Back to the Future

In a time warp wormhole temporal loop nexus bubble reminscent of the worst Star Trek plot-crutch, the rest of Earth caught up with /. today, when Firefox 2.0a1 (sshhhh! Don't call it that) Bon Echo Alpha 1 was actually released.

In a way, I wish the reports had been right: I've been sick for the last three days (breathing is overrated, let me tell you!), so hearing that that the release was done without me having to lift a finger would've been... nice.

Now quick: someone go publish that 1.5.0.2 and 1.0.8 are all out the door, too, so I can get some rest!

March 17, 2006

kswapd has been running for how many DAYS?

Most of the time, our Build Farm makes me a very unhappy camper.

This is because most of the time, I'm interacting with it in a firefighting/rescuing builds from a burning tree building role.

There are a lot of machines in the Farm that are reviled (gaius, anyone?), and they get a lot of attention.

But there are other machines that, for the most part, quietly do their job, day in and day out, and we never hear from them and they become the well-behaved, but unloved and ignored children of the Build Farm.

Today, I was introduced to such a child when I logged into a particular machine doing Linux 1.0.x Firefox builds, and noticed this:

cltbld@madcow tinderbox]$ uptime
6:03pm up 351 days, 23:37, 3 users, load average: 0.13, 0.18, 0.09
With a name like madcow, I wouldn't expect such... stability.

But there you have it: our Build Farm star of the week!

Bits for every endianness and a binary for every instruction set

After some trials and tribulations (and lots of Tinderbox hugging), we finally have universal binaries for 1.5.0.2!

I just noticed they're being uploaded to the wrong location, so I'll have to fix that, but you can find the mixed-endianess here.

When they move, you can find the builds via the tinderbox-with-the-boring-name (and a "UniversalBinaries" tag), on the 1.8.0 Tinderbox page.

Special shout out to Mento and Josh, who helped us (me) tremendously in getting these builds available.

Truth be told, it took a little longer than it otherwise might have because...

...one of the things I'm working on is documenting and recording the reference platforms for "official" Firefox and Mozilla builds, and this was the first platform/machine that I did this for.

"Reference platforms" is a part of configuration management that the Mozilla project hasn't really examined closely for its shipped builds.

I've noticed this when I ask developers what they think the reference platform is or should be. I often get answers like "Well, we need VC8 and... that's it" or "It'll build on Fedora Core 4 or CentOS or RHEL3 or..."

This is an interesting aspect of open source projects in general and Mozilla in particular, in that these are all valid answers if the question is "What platforms should Mozilla build on?"

But in terms of "reference platform," it doesn't answer the question. To support, QA, and ship a set of builds, the answer to what a reference platform is includes the operating system version(s), the patches applied to that installation of the operating system, the entire compilation toolchain (compilers, linkers, system libraries, etc.), any patches/extensions/SDKs to the toolchain, and even the tools used to support the build (Perl, CVS, etc.).

All of these must be recorded and archived as necessary as a fundamental requirement of release engineering. It allows the build to, among other things, be reproduced if necessary.

Mark and Josh were very helpful and patient while they (unwittingly) played guinea pig while I figured out all the details that needed to be recorded. Thanks again to them.

As for developers focusing on the other platforms/branches: I'm coming for you.

March 13, 2006

My own personal www.thedailywtf.com.

So I've been doing some virtual machine migrations—more on that later—but this one relatively old RedHat VM was having problems booting after the migration.

One of the details that has to be taken care of is the physical machines are unlikely to have the same SCSI disk drivers that the VM has, so you have to futz with kernels and modules (oh my!)

While running mkinitrd(8), I kept getting an "/sbin/mkinitrd: line 477: /sbin/tune2fs: Permission denied" error.

I investigated a bit further, and found the permissions on /sbin/tune2fs to be incorrect. After fixing them, mkinitrd spat a more interesting error.

More investigation revealed:

bash-2.05b# file /sbin/tune2fs
/sbin/tune2fs: timezone data
bash-2.05b#
Yeah.

It's been one of those Mondays...

March 11, 2006

On stealing electrons

<rant mode="on">If you own a coffeeshop and you advertise your "FREE WIRELESS" access (with a neon sign, even!), then you need to offer your customers plugs.

Lots of plugs.

Customers shouldn't have skirt the wall of the store, searching for a place to plug in (which, like an oasis in the desert, is typically already being used by everyone else who needs power).

Every table should have a one-to-one mapping of seat-to-outlet.

The only local place I've seen who actually does wireless correctly is Palo Alto's Happy Donuts, which not only has outlets all over the place, but has extension cords and power strips for you to borrow. Unlike most every other place I've seen, they want you to stay there for a few hours and use their wireless, and I for one am happy to pay a quarter or so more per cup of coffee for the pleasure.

This is in direct contrast to most places who do wireless, and just don't care. They don't optimize for offering that service at all, which makes it mostly pointless.

The worst offender is Mountain View's Dana Street Coffee, which has a bunch of signs specifically telling you not to plug in. And I delight in letting everyone I know not to go there, even if they aren't planning on using the wireless.

If they're that rude and incompetant towards their wireless customers, how good can the service/quality of the coffee be?

</rant>

March 8, 2006

On the (continuing) cute surprises of Mozilla lore...

I just checked in the patch for Bug 329686 (which should hopefully make AUS more useful to open source projects, including... y'know, ourselves!) and received this handy dandy reminder of my responsibilities as a developer with Mozilla CVS checkin access:

Date: Wed, 8 Mar 2006 21:47:22 -0800
From: bonsai-daemon@mozilla.org
To: preed@mozilla.com
Subject: [Bonsai] You're on the hook now!

You are responsible to make sure the build of the Seamonkey Tree works.

You just checked into mozilla/config the files: autoconf.mk.in. From now until the tree is verified and opened again, you are to be available for the build team to pester in case of any problems.

I find this amusing for a number of reasons. But most of all, I think, because I'm pretty sure that I know where I'll be for the rest of evening... in case I need to pester myself due to any problems...