Michal Čihař - ColorHug client in unstable

ColorHug client in unstable

It took me much more uploads than I expected, but ColorHug client is finally in unstable. I was fighting with various bugs, which I could not reproduce locally and of course with my own stupidity in configuring pbuilder.

Anyway the buildds overview now shows mostly green, so this is good. The remaining ones mostly wait for libgusb, which needs build on armhf and needs to be ported to non Linux (or rather libusb needs to exist there).

One of things I've learned on the way is that docbook-utils is not enough to convert docbook to man pages, you need docbook package as well. Still I don't understand why this did locally work for me...

The one outstanding issue is that client code assumes it runs on little endian host (or rather same endianity as the ColorHug device) and thus will break on PPC and other big endian machines. The upstream is aware of that an I even managed to produce patch which should fix it, though I have no means to test it (lack of big endian hardware and lack of the ColorHug device itself).


Johannes Berg wrote on Jan. 5, 2012, 5:55 p.m.

I thought upstream committed my original patch for big endian? They might well have broken it again ...

Once I get the device I'll be able to test it on powerpc though.

wrote on Jan. 5, 2012, 11:25 p.m.

The fixed point math still assumes host is little endian (same as device) not sure if there was another issue as well.

foo wrote on Jan. 6, 2012, 2:50 a.m.

How about packaging the bootloader, firmware, schematic and PCB too?

wrote on Jan. 11, 2012, 11:24 a.m.

I don't see much use for these having packaged...