[Nut-upsdev] libusb usage

[Michal Hlavinka]

On Tuesday, December 07, 2010 14:19:59 Charles Lepple wrote:
> On Dec 7, 2010, at 7:10 AM, Arjen de Korte wrote:
> > Citeren Michal Hlavinka <mhlavink at redhat.com>:
> >> I've been asked to find out some answers from libusb maintainer, so
> >> I'm
> >> forwarding them:
> >> 
> >> - Why nut uses legacy libusb-0.1 api and not libusb 1.0 ?
> > 
> > Well, mostly because at the time we started adding USB support to
> > NUT, libusb-0.1 was all there is.
> 
> Also, there is a little more static friction to overcome if we switch
> to libusb-1.0 - in order to keep configure-time complexity down, we
> would want to switch *all* of the drivers to libusb-1.0 (since not all
> of the USB-based drivers are served by the common code that usbhid-ups
> uses). The real trick there is runtime testing of all of the drivers.
> 
> If you are trying to eliminate direct dependencies on the old
> libusb-0.1 package, there is always the libusb-0.1-compat package. I
> believe that FreeBSD 8+ has incorporated that into their libusb-1.0
> API for the new USB stack, and aside from some changes in our
> configuration scripts, it seems to work well at runtime.

AFAIK libusb maintainer works on libusb-compat and wanted to know whether 
it'is be sufficient to place libusb to /lib or libusb-compat is needed in /lib 
too (for nut shutdown and /usr/ unmounted)

> NUT as a whole doesn't reap many of the benefits of libusb-1.0 since
> the NUT architecture uses a single thread and process to interact with
> the hardware. A more integrated system like UPower might be better
> served by the asynchronous calls in 1.0.

thanks for answers