Disapearing pl2303 usb serial adapter on rpi2

[bob prohaska]

On Sat, Dec 23, 2017 at 06:37:29PM +0100, Hans Petter Selasky wrote:
> 
> Idea: Cut a USB cable, reconnect all the wires except the red one, and 
> put a multimeter capable of measuring mA's between.
> 
> --HPS

That thought has crossed my mind, but I'm hesitant at the moment because it
requires a relatively large disturbance to the host and wiring.

The discovery of the USB connector shroud being not soldered to the 
signal ground was a real surprise, now corrected in one instance. That 
adapter has now run 20 hours, in contrast to the last several sessions
which locked up overnight. 

Until the lockup can be reproduced it seems wise to minimize physical 
changes. I suspected a hardware problem at first, but the long (months)
hiatus followed by a recurrence after upgrade to r326951 persuaded me to
suspect software, since the hardware literally hadn't been touched. Now,
again, I'm not so sure.

Many years ago I worked on a development project with a patient software
engineer, and we sparred constantly over the "Is the problem hardware or
software?" question. In that case he wrote a test program that could be left
carefully unchanged. It exercised the hardware we were working on and didn't
touch anything else.. This was 1997, on ISA-bus computers running QNX.

There seems to be an open source hardware test suite called Inquisitor,
https://en.wikipedia.org/wiki/Inquisitor_(hardware_testing_software) 
Is there anything like it in FreeBSD? Peter Holm's stress2 suite
is in the right spirit, but isn't directed at USB or ARM. 

Long-latency problems of the kind I'm seeing are hard to sort out 
under any circumstances. Grounding and shielding issues are the 
worst of the lot. If trouble shows up once a month, a reversion 
to the last working edition covers a lot of changes. Test fixtures
can obscure underlying defects, like unsoldered shields. 

Thanks for your attention, and apologies for what's beginning to
look like  a red herring!

bob prohaska