[Lcdproc] Crystalfontz 633/635 Packet Driver Endianness Issue

[Russell Krayer]

Here is an update ...

** LCDd on ppc with lcdproc on ppc
The backlight turns on and off randomly
Remnants of different screens remain after switching to new screens

** LCDd on AMD64 with lcdproc on ppc
The backlight "flickers" rapidly
Screens seem to work fine otherwise

** LCDd on ppc with lcdproc on AMD64
Initailly the backlight was off but turned on at one point and remained on
Screens all seem to work fine

** LCDd on AMD64 with lcdproc on AMD64
Everything works fine


Any suggestions?  Testing to this point seem to suggest problems with 
lcdproc on ppc.


Thanks!
Russ


>From: "Russell Krayer" <rkrayer@msn.com>
>To: lcdproc@lists.omnipotent.net
>Subject: Re: [Lcdproc] Crystalfontz 633/635 Packet Driver Endianness Issue
>Date: Sun, 10 Sep 2006 13:39:08 -0400
>
>Hi Peter,
>
>Your fix in the nightly build seems to work - thanks for the quick 
>response!
>
>But now I think I am noticing other problems --- I will post again when I 
>can characterize it better.
>
>Regards,
>Russ
>
>
>>From: Peter Marschall <peter@adpm.de>
>>To: lcdproc@lists.omnipotent.net
>>CC: "Russell Krayer" <rkrayer@msn.com>
>>Subject: Re: [Lcdproc] Crystalfontz 633/635 Packet Driver Endianness Issue
>>Date: Sat, 9 Sep 2006 22:18:55 +0200
>>
>>Hi Russell,
>>
>>On Saturday, 9. September 2006 17:10, Russell Krayer wrote:
>> > I stumbled onto an issue with endianness running LCDproc 0.5.0 on a ppc 
>>box
>> > (a big endian machine).  The problem is in CFontz633io.c when working 
>>with
>> > the CRC.  I have a fix that is working for me and should still work 
>>with
>> > little endian machines.  Should I send a patch somewhere?  Note, the 
>>bug is
>> > also in the Crystalfontz example program from Crystalfontz vendor.
>>
>>Thanks for the patch you sent in your other post.
>>
>>I looked at it and the existing code and finally found that the idea of
>>using a union for this conversion was a flawed concept.
>>
>>So I changed the code in CVS to convert manually between the
>>CRC and the 2 bytes comprising it.
>>This should fix the endiannes issue as the code now always expects / sends
>>the LCB before the MSB.
>>
>>As a side note: unfortunately CF uses little-endian word order here.
>>With big-endian, a simple htons() had been sufficient.
>>
>>The changes should be in tonight's nightly tar balls at
>>http://lcdproc.sourceforge.net/nightly/lcdproc-CVS-current.tar.gz
>>
>>Please test if the changes work for you
>>
>>Peter
>>
>>--
>>Peter Marschall
>>peter@adpm.de
>>_______________________________________________
>>LCDproc mailing list
>>LCDproc@lists.omnipotent.net
>>http://lists.omnipotent.net/mailman/listinfo/lcdproc
>
>
>_______________________________________________
>LCDproc mailing list
>LCDproc@lists.omnipotent.net
>http://lists.omnipotent.net/mailman/listinfo/lcdproc