[sane-devel] Problem with libusb and 64 bits 2.6.25 kernel

Nicolas nicolas.martin at freesurf.fr
Tue May 27 20:50:29 UTC 2008


Le mardi 27 mai 2008 à 16:24 -0400, m. allan noah a écrit :
> On Tue, May 27, 2008 at 4:01 PM, Nicolas <nicolas.martin at freesurf.fr> wrote:
> >
> > Le mardi 27 mai 2008 à 15:42 -0400, m. allan noah a écrit :
> >> On Tue, May 27, 2008 at 3:29 PM, Nicolas <nicolas.martin at freesurf.fr> wrote:
> >> > Le lundi 26 mai 2008 à 19:12 -0400, m. allan noah a écrit :
> >> >> two random thoughts-
> >> >>
> >> >> 1. can he try 32 bit kernel on same exact machine, just to rule out
> >> >> hardware problems?
> >> >
> >> > Good point, and to try other USB ports on the same machine, that needs
> >> > to be confirmed first as it could be a HW issue
> >> >
> >> >> 2. is there a pattern to the timeouts, like always same number of
> >> >> errors before a good packet?
> >> >
> >> > It looks to be a repetitive pattern, i.e. error occurs on a second
> >> > "read" sequence when reading image data, but occurs also sometimes with
> >> > smaller transaction control messages (write message, then read response
> >> > immediately).
> >> > I've asked to give a try with a small tempo between two successive image
> >> > data readings, but do not have feedback yet.
> >> >
> >> > Thanks
> >> >
> >>
> >> yes- it could be that his machine is too fast for the scanner, not
> >> that it is 64 bits.
> >>
> >> allan
> >
> > But this would mean that with machines getting faster and faster,
> > something needs to be adjusted, so that consecutive usb commands do not
> > occur within a too short delay.
> >
> > Unfortunately, I did not see any spec for such timings in libusb.
> > Maybe something about that is the usb spec ? need to check this point...
> >
> > Nicolas
> >
> 
> there is no way to control this in libusb AFAIK, instead, you would
> have to determine a dynamic delay period to add before all commands.
> 
> allan

Probably will propose a static delay, even if that does not optimize
transfer speed performance (is that important for a scanner anyway ?),
it will be IMHO, more robust.

Nicolas





More information about the sane-devel mailing list