Pieter Nagel <[EMAIL PROTECTED]> wrote: [...] > - A workaround is to modify usb/scanner.h and change IBUF_SIZE to be 64k > instead of 32k. Why this works I do not know.
Thanks for the tip. I will try this ASAP. > - A surefire way of reproducing the bug is to do any scan whose width is > 2870 pixels wide, at 2900 dpi. Scan height has no effect, which seems > consistent with the fact that this is a data transfer problem - the > wider the scanline, the more data to transfer. It _is_ indeed a data transfer problem. But the Ğmore datağ part is not the whole story, at least not in a mere dpi context. I can reproduce the problem 100% also in a _prescan_ mode of Sane, with a very modest resolution (~200dpi, don't remember the exact figure). When I _increase_ the prescan resolution somewhat, it works OK! But then, I do not know exactly the correspondance between resolution and amount of data transmitted as a function of time, which is obviously the crucial part if a buffer overflow is involved. This factor I would assume to be strongly application program dependant. > How to proceed from here? I see two possible approaches: Either attract the attention and interest of somebody who knows the code, or dive into the source ourselves and eventually submit a nice patch. Being Open Source, the latter _is_ possible. ;-) -- Jon ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
