[sane-devel] CanoScan LiDE 220 works, but only with debug enabled

[mark at fastmail.com]

On Mon, Apr 6, 2015, at 07:06 AM, Michael Shigorin wrote:
> On Mon, Apr 06, 2015 at 02:15:27PM +0200, Stef wrote:
> > >I tried turning on debug to find out what is happening and found that
> > >everything works perfectly if I enable massive debugging:
> > Since the usual way to find out bug is to enable debug traces,
> > it is going to be difficult to solve your problem.
> 
> Might be race condition if enabling debug changes timings a bit;
> in that case putting some load on *all* CPU cores (like make -j
> upon something large or enough yes(1) processes stuck into
> /dev/null) might change timings over the same margin; but that's
> about reproducing "with debug" shift, not about debugging without
> masking the problem.
> 
I pulled git code again and rebuilt on an older, slower machine with
only USB2 interfaces and it works perfectly. The machine with the
problem has only USB3 ports. I noticed some list traffic recently about
a kernel bug in USB3 and wonder if this is a manifestation of that. I am
running 3.19.0-1.slh.2-aptosid-amd64 as my current kernel.

Just another data point: If I redirect the debug to /dev/null then the
problem persists. Probably useless information.

-- Mark