http://bugs.winehq.org/show_bug.cgi?id=4055
------- Additional Comments From Germain.Morbe@web.de 2006-17-03 14:59 ------- Meanwhile, i had access to a ICD2 device for about one week and i made several experiments up to the hight of my skill level. To make it short, i did not manage to get the ICD2 working. However, i tried to track down this FIFO issue, which after all is the true reason for this bug in my opinion. John, you are right that there is no way to set the FIFO size under Linux with one exception. You can fool the kernel by setting the UART chip to 8250 which has no FIFOs at all. This works fine with 2.4xx kernels according to here:
http://www.ussg.iu.edu/hypermail/linux/kernel/0408.0/0798.html
As far as they pointed out there, this does not work with 2.6 kernels because of a bug in the serial module of the 2.6 kernel. And as far as i could make clear this is very sad but never the less true.
Simply using two small terminal.exe apps, one under wine and another one under windows, both connected via a nullmodem cable it can made sure that the kernel 2.6 when set to UART 8250 sends out only one character out of 16. So 15 characters are dropped to nirvana instead of sent. Aafter all im sure that would you use an old serial card with a real 8250 chip under a 2.4xx linux kernel you could probably connect to and work with your ICD2 like a charme.
However, since my Picstart as well as our Promate II works well under wine, i can only wonder why those Microchip idi*ts choosed to use a completely different port setup for the ICD2 communication.
Finally i wonder if this "bug" ever can be fixed within wine or if this is a more general linux 2.6 bug which can only be fixed by the kernel guys. I would be glad to hear tony's opinion about this, thanks.