I can agree as well, the problem does not exist on my flashme'd DS, just on the DSL; aka: "
Maybe they've not seen the problem in person, could be model/chip specific." At least I'm 100% right on that point, anyway. That's always been my biggest frustration with libnds as well, they always seem to try to break *everything* on updates
I swear, I've spent 5x more time looking at libnds source updates than I have coding when it comes to the DS.
PersonalData's data it stored on a spi accessed firmware chip, it's
read from there before you have access to the info (and it's just as much if not more of a pain to put that data back, it takes a clear chunk of code to do so), be it by the bios or otherwise. To change the brightness setting, well
here (which was why I suggested it may be the brightness code, as that method tries to write it back to the user data - wasn't there a couple reports of losing the language setting as well?) I do recall there being problems surrounding PersonalData corruption before, back around DKPR17 or so, but I think it was only with that asian one with the larger firmware chip. Too bad they used CVS for so long, the commit logs are so much better in SVN.
Would it help any if I coded up something to dump both slots of PersonalData before and after the problem comes up? As far as I can tell that would only be proof rather than a fix - are you then (basically) using gelu's codebase right now for akmenu arm7? If so (and I assume not, more likely what is in AK-BBS) I'll check out what is current and see if I can replicate the bad result on an independent/small app. Anyway, if there is anything I can help with in isolating it, I'm still game.
BTW, I did find NDSX_ARM7_FirmwareAccessFifo, in google at any rate - in lick's DSLua stuff... which led me to look a little more closely at gelu's .h files - what a mess putting code in a header