Finally had something to update on my project page, listed off site.

You can check it out on http://www.nec.fct.unesp.br/marcos_umino/soc.html

And I hope to see all of you here by the end of MidTerm Evaluation!

Cheers!

Yesterday I finally managed to flash GPE on my PDA. But, the problems just begun to appear.

Even with Windows Mobile installed I had problems when usign USB connection - with WinXP it didn't even identify device, in linux it at least listed the device correctly, although the ipaq.ko module, which takes care about managing ttyUSB device needed for communication, did not work - strange thing was that on other computers it worked correctly with this same ipaq.

So, I hoped that GPE image will help me to solve these problems, but it appears that this is not true at all. With GPE image, the device is not recognized correctly in linux nor windows. It seems that my USB controller doesn't like the ipaq. And, as if it was not enough - my laptop's wireless adapter doesn't have any native linux drivers, so I have to run ndiswrapper. The windows NDIS driver does not handle ad-hoc networking correctly, not mentioning that the AP mode was quite strange enough. Not having bluetooth adapter nor serial port on laptop, the only way for communication appears for now the IrDA.

This is bit OT post, but not too OT - it is a related to preparations of development environment for PDA X server emulation. Few weeks ago, when I started a bit playing with window manager for PDAs, matchbox, I faced a strange bug, which caused matchbox crashes. This is in general quite strange, because on 32bit platforms matchbox works without any similar problems. And, browsing code of matchbox did not bring anything but few bits of confusion - everything seemed properly. So, after few approaches to understand the problem, I gave up.

Recently, the real cause got revealed - problem was that some of X functions (namely XChangeProperty and XGetWindowProperty). They are used to set/retrieve some additional informations associated with X windows. One of their arguments is 'format', which (on 32bit machines) says the size of one 'item' stored in property. The functions then expect array of these items. But, there is an important phrase in manpage, which was overlooked: " If the returned format is 32, the returned data is represented as a long array...". And the important thing is, that on x86_64, size of long is not 32, but 64 bits. The function expects array of 'long'-s, from which it always takes lower 32bit and discards the rest. So the resulted data which were stored and consecutively retrieved with old method, missed every odd 'int'. And, with data damaged this way, matchbox could crash.

Since I have not had much time for any activity except school and few things around, my first blog post comes a bit "late" :) To quickly summarize: I will try to blog about my SoC project - GPE Network Configuration application, and, of course, related things.

Funny thing is that I have never written any blog, so quality of this one may vary greatly. Stay tuned for more info, it will show up everytime I will have time to write. And much more often after finishing my finals:)