Shining the Chrome

Reflected in Chrome

Chrome apps can exist as pages or panels. So far I’ve been testing Disk Clock primarily as a panel app, because it seems too small for a full page. However, the panel also seems a little awkward (Opera widgets do better here, allowing transparency and a small floating set of buttons) Ultimately, I need to take the final steps to full scalability, allowing Disk Clock to be nearly full screen on ChromeOS devices (and others, of course)

Chrome Packaging

Disk Clock is now running as a packaged Chrome app from a locally built package file. Previously it was running as an “unpacked extension”, which only requires manifest.json and a few icons. (If I put it in the store, I’ll just need a zip file, and Google will build the package.)

Packaging can be done as a GUI operation from the extensions page. Importantly, for my multi-clock build system, it can also be done on the command line. To set up, you have to run a package through the GUI once; leave the key file out and Chrome will generate one. Thereafter, it can be used to specify that this is the ‘same’ package, perhaps in a version. The command line build can take the key as an argument, making it possible to fully automate the process past the first time.

Build Graphing

A small piece of the DC build dependency graph

Building the various packages takes some time (mostly the RequireJS compiling part), so refactoring my Rake based build system was a potentially tedious process. Rather than test by building, I pulled in GraphvizR and generated Graphviz dependency diagrams to check things out before trying an actual build. Because I’m building so many packages, with so many shared dependencies, I had to limit output to a single target at a time in order keep it comprehensible.

Attacking the Clones

When I first installed all the dashboard widgets to verify the packaging, there was a confusion of identical icons. I finally have icon formatting and sizing built into the build system by way of RMagick Grabbing screenshots is still a manual process, and will probably remain so for some time – I would probably have to do a special build with some sort of startup script and then launch a browser to automate the actual capture.

The Latest Fashions

I wasn’t really happy with the half-and-half color scheme that I’ve let lie for some time. Initially I went back to the old standby quarters, but I had a longing for the stronger sense of position one gets from the day disk. I’ve made two further experiments. One, directly imitating a sample proportion of the day disk with more of one color than the other, allowing it to better indicate the root mark. Two, I’m using two shades of the color instead of the color and stark white, which was a little to jarring. Hopefully this will make each disk a little more cohesive.

No Roads on the Moon

Calendar set, showing unmarked moon.

Another small change I’ve toyed with for some time is getting rid of the marker lines on the moon disk. Perhaps it was just my clumsy art, but it seemed a little jarring. Given that I’m not simulating all the physics involved and it can be off by about fourteen hours, precise marks seem kind of out of place anyway.

Posted Friday, February 18th, 2011 under Devlog.

Comments are closed.