Re: [Qemu-devel] Should we auto-generate IDs?

[Programmingkid]

On Sep 3, 2015, at 10:43 AM, Jeff Cody wrote:

> On Thu, Sep 03, 2015 at 10:34:04AM -0400, Programmingkid wrote:
>> It has been over a week since we first started talking about this
>> subject. A lot of opinions have been flying around. Does this issue
>> look like it is starting to be fixed? I just have to say I don't
>> think it has been solved yet. Having device_del use a QOM path does
>> not sound very good. It is actually easier and faster for the user
>> to restart QEMU with a new ID for some device than to have to look
>> up some QOM path. Having a user-friendly system for creating ID's
>> should be a priority. 
> 
> Please see my v2 patch series (you were CC'ed on it):
> 
> 
>    [PATCH v2 1/2] util - add automated ID generation utility
>    [PATCH v2 2/2] block: auto-generated node-names
> 
> The first patch is a qemu-wide utility to generate IDs.
> 
> The scheme seems to have some consensus, and a few r-b's (it is
> also derived from this conversation thread).
> 
> The second patch is using it in the block layer, for node-names; it
> should be trivial to see how to use it for qdev.
> 
> Would you like me to roll a v3, with a qdev patch added in?
> 
> Thanks,
> Jeff

Yes, thank you for them. The first patch does provide a function that generates 
an ID, 
but it doesn't actually give the ID to anything. A third patch might be needed
that actually puts the id_generate() function to use. I use USB devices that 
I would like to be able to remove during QEMU's usage. Any ID generating
system would be very useful. In qdev-monitor.c there is a function called
qdev_device_add(). That is where I would use your id_generate() function.