Hi,

On 06/12/2013 01:45 AM, Pete Batard wrote:
> On 2013.06.11 11:10, Hans de Goede wrote:
>> I think it would be good to tag current master as 1.0.16-rc1, spin
>> tarbals, etc.
>
> I'd prefer if we don't tag the RC-1 until next Monday, on account that:
> - I'd like to keep at least a little distance between 2 major events,
> such as the announcement of libusbx and libusb merging back and libusbx
> getting into release mode. There's the possibility that, since libusbx
> is planned to be the effective continuation of libusb, people over at
> libusb may request a few things from the next libusbx release.
> - It's nice to give people a few days before RC gets into effect, in
> case they are working on something they'd like to see integrated, before
> we freeze the addon of new features.
> - It'll give us exactly 2 weeks of RC time to the officially planned
> release, as per our milestones, which I think is what we agreed on for RCs.
>

Ok, that works for me. I still would like to see one more feature added
to 1.0.16, and this will give me this for that :)

I'm talking about adding a libusb_enable_auto_detach_kernel_driver function
and implementation for Linux. Rationale:

1) I believe this is what most apps will really want (libusb taking care of
the kernel driver management / detach + re-attach).

2) Linux has a new ioctl now to do detach + claim-interface in one go in a
race-free manner. We've discussed exporting this before and came up with a
different solution, but now I believe this is a better way of exporting this.

3) This will make it much easier for cross-platform code to deal with kernel
driver detach / re-attach. Instead of having compile time (or now with
capabilities runtime) checks for this, and then needing to error-check the
detach result on supported platform, they can simply do:

libusb_enable_auto_detach_kernel_driver(device);

This which will return either LIBUSB_ERROR_NOT_SUPPORTED or LIBUSB_SUCCESS,
so most apps can simply ignore its return value, since on platforms where
this is not supported they won't care about detaching / re-attach anyways.

I don't believe this is worth slipping the 1.0.16 schedule for, so I'll try
to write an RFC patch for this tomorrow. Then people can review over the
weekend I can push it on Monday. Or if I fail to find time / the patch fails
review in a major way, we can punt this to 1.0.17.



>> I'm willing to the work for this (*), with the exception of providing
>> windows binaries.
>
> Sounds good to me. Thanks.
>
>> Pete, I believe that you've a checklist for the release process somewhere ?
>
> Hidden in plain sight, at
> https://github.com/libusbx/libusbx/wiki/Maintainers%27-Corner#wiki-RCRelease_work

Thanks for the pointer.

Regards,

Hans

------------------------------------------------------------------------------
This SF.net email is sponsored by Windows:

Build for Windows Store.

http://p.sf.net/sfu/windows-dev2dev
_______________________________________________
libusbx-devel mailing list
libusbx-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libusbx-devel

Reply via email to