Re: [dpdk-dev] [PATCH] devtools: check stable tag in fixes

[Thomas Monjalon <thomas.monjalon@6wind.com>]

2017-01-18 12:41, Yuanhan Liu:
> On Tue, Jan 17, 2017 at 07:42:33PM +0100, Thomas Monjalon wrote:
> > 2017-01-17 18:15, Ferruh Yigit:
> > > On 1/17/2017 2:54 PM, Thomas Monjalon wrote:
> > > > The tag "Cc: stable@dpdk.org" must be set when the commit must be
> > > > backported to a stable branch.
> > > > 
> > > > It must be located just below the "Fixes:" tag (without blank line)
> > > > and followed by a blank line, separated from SoB and review tags below.
> > > 
> > > I am OK to keep it if it will help stable tree maintenance, but I still
> > > not clear about why we need this.
> > > 
> > > If a patch is a fix, it should already have "Fixes:" line, so this can
> > > be used to parse git history.
> 
> Same answer (as I have already replied to you in another email): not all fix
> patches should be picked to stable branch. (I gave some examples below)
> 
> > That's a question for Yuanhan. My comments below:
> > 
> > Some fixes are not candidate for the stable branch because the bug was
> > not in the previous release. These fixes are filtered out by the script
> > devtools/git-log-fixes.sh
> > Some fixes are not so important and we can decide they do not fit in
> > the stable branch.
> 
> Yes, I see no reason to pick patches that fix a typo, a comment issue,
> a coding style issue, or even, a hypothetic bug.
> 
> Usually, it should be a patch that fixes a solid bug, like a crash, or
> something like "it behaves abnormally without the fix". That's the basic
> rule. Upon that, I think we could lower the bar a bit case by case. For
> example, I picked a doc update to v16.07.2:
> 
>     commit 92b70d21ee29bad92766699d0b45f579a2ff9adc
>     Author: Jingjing Wu <jingjing.wu@intel.com>
>     Date:   Fri Sep 30 14:46:23 2016 +0800
>     
>         doc: add limitations for i40e PMD
>     
>         [ upstream commit 972cc03ac4e30a7df8f734a77021bb15d0419b55 ]
>     
>         This patch adds "Limitations or Known issues" section for
>         i40e PMD, including two items:
>         1. MPLS packet classification on X710/XL710
>         2. 16 Byte Descriptor cannot be used on DPDK VF
>         3. Link down with i40e kernel driver after DPDK application exist
>     
>         Signed-off-by: Jingjing Wu <jingjing.wu@intel.com>
>         Acked-by: John McNamara <john.mcnamara@intel.com>
>     
> It adds some limitations to the code introduced in last release: I think
> it's an important note the user might want to know if he sticks to that
> stable release.
> 
> Talking about that, I think this patch makes abuse of the "Cc: stable"
> tag if a developer has to (or must) add such tag for a fix patch, no
> matter what it fixes.
> 
> > Who make this decision? Relying on this Cc tag would
> > mean the committers decide which patch to backport.
> 
> Ideally, it's the developer to make such decision: he knows the best what
> he is trying to fix, he also knows which version is vulnerable.
> 
> But that's not something a new contributor might be aware of, and that's
> what the maintainer (not the committer) could be helpful here: tell him
> it's a candidate for a stable release and guide him on howto.
> 
> The reason I want to stress the point of "the maintainer but not the
> committer" is, normally, the maintainer knows the code better.
> 
> > > If patch is a feature, as far as I know still can go to stable tree, but
> > > for this case stable tree maintainer decides this, and author putting
> > > "Cc: stable@dpdk.org" tag not so useful. Author can put this tag just
> > > for recommendation, but if so why we are saving this into git history?
> > 
> > No feature should be backported.
> > 
> > > Initially this was to be sure fixes CC'ed to stable mail list, now
> > > meaning is changing I guess. For the case author already cc'ed the
> > > stable tree but not put Cc: tag into git commit, should committer add
> > > this explicitly or ask from author a new version of the patch?
> > 
> > Yuanhan was suggesting that the committer can do it if an author forget.
> 
> Yes, it's just part of the committer job, say, adding Tested-by,
> Reviewed-by, that kind of stuff.

So it should be stressed in the contribution guide that it is the
responsibility of the author and the maintainer to put this Cc: tag.

> > > Last thing, if this tag will remain in the commit log, is this only for
> > > stable tree, or any "Cc: <mail_address>" can stay in the history?
> > 
> > I do not see the benefit of keeping other Cc in the history.
> 
> Again, I really don't know why you bother to remove it, manually, commit
> by commit. What's the gain here? It just adds more burden to a committer.
> Honestly, I never do that.

I had the personal feeling that if the Cc: person is not in SoB, Ack or
Review tags, it means he's not interested in this patch.
After thinking more, I was probably wrong.

> It actually has benefits. Keeping the cc tags allows us to cc them when
> we find a bug in that patch and make a fix patch later: they are likely
> to want to know any follow updates on that original patch.
> 
> It also helps when I send out a stable patch review: I send a copy to
> all guys on the Cc list, on the Ack/Review, SoB list. I think they might
> also want to know that patch is a candidate for a specific branch. He may
> even give some comments, something like "if you pick this one, you might
> want to pick another one", or "why bother to port it to a stable release",
> that kind of thing.

OK Yuanhan, thanks for the explanations.

I will send a v2 to relax the constraint of blank line below Fixes: tag,
and change the warning when Cc: tag is missing. Or remove the warning?
What do you think of: "Reminder: is it a candidate for stable@dpdk.org backport?"