On Sep 11, 2016, at 1:17 PM, Paul Eggleton <paul.eggle...@linux.intel.com>
wrote:
>
> Hi Jianxin,
>
> On Fri, 09 Sep 2016 10:50:50 Jianxun Zhang wrote:
>>> On Sep 9, 2016, at 7:34 AM, Saul Wold <s...@linux.intel.com> wrote:
>>> On Thu, 2016-09-08 at 17:02 -0700, Jianxun Zhang wrote:
>>>> This change update intel-gpu-tools to 1.15 as a sync-up with
>>>> Intel graphic stack 2016 Q2 release.
>>>>
>>>> This change explicitly sets ${PV} in recipe and renames it
>>>> after package name only. By doing so we don't need to enforce
>>>> a policy to rename recipe every time we do update. Patch
>>>> speaks itself.
>>>
>>> This is wrong!
>>>
>>> The whole point of naming the recipe as ${PN}_${PV} is to remove the
>>> need to update PV in the recipe.
>>>
>>> Yes checksums need to be updated on a regular basis, there are tools
>>> like devtool that can assist with this.
>>>
>>> The OpenEmbedded standard is for $PN_$PV.bb filename this also assists
>>> in understand quickly what version a particular recipe is. There are
>>> also limited cases where 2 different $PVs are needed for some reasons.
>>>
>>> Please resubmit this as a normal update with $PV contained in the file
>>> name.
>>>
>>> Please see: https://wiki.yoctoproject.org/wiki/Best_Known_Methods_(BKMs
>>> )_for_Package_Updating
>>
>> Saul,
>> I think the drawback of current procedure is it causes git “resets" history
>> on recipe since new recipe when a renamed or git mv-ed recipe also have too
>> much modifications. I don’t think git really tracks moving or renaming:
>
> It doesn't reset history. You're right that it doesn't track renames - that's
> because rename detection is done whenever git looks at a change, not when the
> change is applied. If you want to see the full history including renames, use
> the --follow option.
Paul,
I just come back from a vacation and run "git log --follow new_renamed_file" in
the test project created with procedure in my previous reply. It doesn’t show
history before renaming the file in that case.
The procedure simulates a routine update (renaming & modifying recipe in a
single commit) but won’t always give the expected result in git history. I
think the first half of your comment could be the answer. git doesn’t track
rename, so --follow should not have more magic in this case.
>
> Saul is correct, use of PV in the recipe filename is standard OE practice and
> we don't want to be deviating from that - besides which I'm not sure we have
> much to gain by doing so.
If you defined this as a standard in OE universe, no more argument is needed. I
just want to point out any inconsistent outcome from the standard practice
which may be out of expectation sometimes, assuming my test is valid.
Last time I followed the policy to use git mv and submit a patch to update a
short recipe in meta-intel, we had same confusion when Saul didn’t see renaming
info in patch in review. (I can’t remember which recipe however).
I will submit a V2 back to normal.
Thanks lot!
>
> Cheers,
> Paul
>
> --
>
> Paul Eggleton
> Intel Open Source Technology Centre
--
_______________________________________________
meta-intel mailing list
meta-intel@yoctoproject.org
https://lists.yoctoproject.org/listinfo/meta-intel
