[Libosinfo] [libosinfo PATCH] os: Add kernel-url-argument API

[Cole Robinson]

On 5/3/19 8:56 AM, Fabiano Fidêncio wrote:
> On Fri, 2019-05-03 at 13:07 +0200, Fabiano Fidêncio wrote:
>> On Thu, May 2, 2019 at 4:39 PM Fabiano Fidêncio <fidencio at redhat.com>
>> wrote:
>>> Tree based installations will require a kernel-url-argument to be
>>> assed. Let's provide an API to the apps to get this info from
>>> OsinfoOS.
>>>
>>> https://gitlab.com/libosinfo/osinfo-db/issues/28
>>>
>>> Signed-off-by: Fabiano Fidêncio <fidencio at redhat.com>
>>
>> I've discusses this with Cole off-list and we've decided that going
>> for an "inheritance" (pero no mucho) way would be better.
>> I'll change the implementation for something towards the line of:
>> current.kernel_url_argument = current.kernel_url_argument or
>> derived_from.kernel_url_argument
>>
>> With that we'll be able to keep kernel_url_argument as a single one
>> and just replace it in case a new one is added (which will be needed
>> for Fedora >= 19, for instance).
>>
>> A v3 (of the whole thing) will be submitted.
> 
> Okay, I have this work done but before submit the v3, I'd like to raise
> one possible issue and confirm that having this kind of duplication is
> fine.
> 
> In the v2, I've been using as part of the scripts something like:
> <xsl:value-if select="os/kernel-url-argument"/>
> 
> Once we start inheriting kernel-url-argument, the kind of call shown
> above wouldn't be possible anymore, leading us to have, as part of the
> scripts, something like (for RHEL):
> <xsl:choose>
>   <xsl:when test="os/version < 7">
>     <xsl:text> method</xsl:text>
>   </xsl:when>
>   <xsl:otherwise>
>     <xsl:text> inst.repo</xsl:text>
>   </xsl:otherwise>
> </xsl:choose>
> 
> This is duplication of info will be needed because not all the <os/>
> entris will have the <kernel-url-argument/> attribute.
> 
> IMO it still looks better in this way (please, check v1[0]) than having
> to have the entry added in all OSes.
> 
> Cole, what do you think?
> 

Hmm yeah that's a bit unfortunate, but I think it's nicer than the
kernel-url-argument duplication, so thumbs up from me

Thanks,
Cole