<br><br>On Friday, June 29, 2012, Bruce Ashfield <<a href="mailto:bruce.ashfield@windriver.com">bruce.ashfield@windriver.com</a>> wrote:<br>> On 12-06-29 09:36 AM, Darren Hart wrote:<br>>><br>>><br>>> On 06/29/2012 06:15 AM, Bruce Ashfield wrote:<br>
>>><br>>>> On 12-06-29 03:22 AM, Khem Raj wrote:<br>>>>><br>>>>> On Thu, Jun 28, 2012 at 11:14 PM, Darren Hart<<a href="mailto:dvhart@linux.intel.com">dvhart@linux.intel.com</a>> wrote:<br>
>>>>><br>>>>>><br>>>>>> On 06/28/2012 11:04 PM, Khem Raj wrote:<br>>>>>>><br>>>>>>> On Thu, Jun 28, 2012 at 10:50 PM, Darren Hart<<a href="mailto:dvhart@linux.intel.com">dvhart@linux.intel.com</a>> wrote:<br>
>>>>>>><br>>>>>>>> However, I don't have a strong opinion here, I'll happily call it<br>>>>>>>> kernel-devel if that is strongly preferred by some.<br>
>>>>>><br>>>>>>> in oe terminology it will be -dev<br>>>>>><br>>>>>> THe other reason I didn't go that route is that we already have a<br>>>>>> kernel-dev (although I don't much like the way it is used):<br>
>>>>><br>>>>>> From kernel.bbclass:<br>>>>>><br>>>>>> PACKAGES = "kernel kernel-base kernel-vmlinux kernel-image kernel-dev \<br>>>>>> kernel-misc kernel-headers"<br>
>>>>> FILES = ""<br>>>>>> FILES_kernel-image = "/boot/${KERNEL_IMAGETYPE}*"<br>>>>>> FILES_kernel-dev = "/boot/System.map* /boot/Module.symvers* /boot/config*"<br>
>>>>> FILES_kernel-vmlinux = "/boot/vmlinux*"<br>>>>>> # misc is a package to contain files we need in staging<br>>>>>> FILES_kernel-misc = "/kernel/include/config /kernel/scripts<br>
>>>>> /kernel/drivers/crypto /kernel/drivers/media"<br>>>>>> FILES_kernel-headers = "/usr/src/kernel-headers"<br>>>>>><br>>>>>> And looking at that it appears I broke kernel-misc by moving things to<br>
>>>>> /usr/src (didn't notice that in my testing).<br>>>>>><br>>>>>> How would you propose we redefine the above in order to use "kernel-dev"<br>>>>>> for the new package?<br>
>>>><br>>>>> hmmm interesting so I guess, linux-dev being already taken, you might<br>>>>> call it linux-kernel-headers as you were doing. So we have<br>>>>> linux-libc-headers and linux-kernel-headers to differentiate between<br>
>>>> raw and sanitized headers<br>>>><br>>>> What would break if the new files were simply added to the existing<br>>>> kernel-dev<br>>>> package ? I'm already putting System.map and others on targets in<br>
>>> different environments for dev/debug, so would adding enough to build<br>>>> kernel modules on the target be a big problem ?<br>>><br>>><br>>> I was considering this last night as well.<br>
>><br>>>> But I probably just don't understand *what* the existing -dev package<br>>>> is used for, and my comment makes no sense :)<br>>><br>>> And I landed here as well.<br>><br>
> :)<br>><br>>><br>>> Should we attempt this approach and only break them apart if somebody<br>>> screams?<br>><br>> In the absence of definitely proof that we are missing something, I'm<br>
> ok with suggesting that we have a single -dev package.<br>><br>>><br>>>><br>>>> Outside of that, I also like linux-kernel-headers as the package name.<br>>><br>>> Would you prefer this even if we merge it with the existing kernel-dev?<br>
>> I think my order of preference would be:<br>>><br>>> 1) Merge with kernel-dev and leave the name as kernel-dev for<br>>> consistency<br>>><br>>> 2) New package called linux-kernel-headers<br>
><br>> Is there a 3rd option of a virtual package called linux-kernel-headers<br>> that maps to -dev ? I have no idea if that is even possible or desirable,<br>> but I thought I'd mention it.<br>><br>> I like #1, since we really are talking about development vs just some<br>
> headers lying around :)<br>><br><br>I think packaging more into -dev is OK upgrade wise so let's use kernel-dev and.put it in usr/src/kernel<br>> Cheers,<br>><br>> Bruce<br>><br>>><br>>> --<br>
>> Darren<br>>><br>>>><br>>>> Cheers,<br>>>><br>>>> Bruce<br>>>><br>>>>><br>>>>>><br>>>>>> --<br>>>>>> Darren Hart<br>
>>>>> Intel Open Source Technology Center<br>>>>>> Yocto Project - Linux Kernel<br>>>>>><br>>>>>><br>>>><br>>><br>><br>>