On Wed, Apr 01, 2015 at 03:30:50PM +0200, Michal Privoznik wrote:
On 01.04.2015 15:02, Martin Kletzander wrote:
> On Wed, Apr 01, 2015 at 11:01:11AM +0200, Michal Privoznik wrote:
>> The 7c3c7f217ebae5 commit introduced a nodeinfo test. In order to do
>> that, some parts of sysfs had to be copied. However, sysfs is full of
>> symlinks, so during copying some symlinks broke. Remove them, as on
>> different systems they can point to different files or be broken.
>>
>
> That commit introduced only two thirds of the files, the rest is from
> f5c2d6 (just in case you want to update the commit message).
>
> But! Not only we don't need these, but power and uevent don't need to
> be there either, do they? And core_siblings(_list) is also not
> something we (will) use since it's just a syntax sugar for parsing
> data for all CPUs (that we do anyway).
>
> If you want to clean the test data a bit, I suggest you clean
>
tests/nodeinfodata/*/cpu/cpu*/{uevent,power,topology/core_siblings*,firmware_node,subsystem}
>
> files too. That's 202 files you can get rid of.
good point. Although I'd leave core_siblings and topology - it's
currently not used, but it may come handy when debugging our own node
info code. It contains all the pieces of information that one need to
reconstruct the topology. But others can be removed as you say. Will
send v2.
I meant that the info in core_siblings* can be reconstructed from all
the other data, but that's right that we might validate it with that
additional files.