Re: [dpdk-dev] 19.11.4 patches review and test

[Thomas Monjalon <thomas@monjalon.net>]

22/03/2021 12:59, Luca Boccassi:
> On Mon, 2021-03-22 at 11:41 +0000, Bruce Richardson wrote:
> > On Mon, Mar 22, 2021 at 10:49:54AM +0100, Christian Ehrhardt wrote:
> > > On Thu, Mar 18, 2021 at 7:25 PM Pai G, Sunil <sunil.pai.g@intel.com> wrote:
> > > > Hi Christian, Ilya
> > > > From: Ilya Maximets <i.maximets@ovn.org>
> > > > > On 3/18/21 2:36 PM, Pai G, Sunil wrote:
> > > > > > Hey Christian,
> > > > > > 
> > > > > > <snipped>
> > > > > > 
> > > > > > > back  in 19.11.4 these DPDK changes were not picked up as they have
> > > > > > > broken builds as discussed here.
> > > > > > > Later on the communication was that all this works fine now and
> > > > > > > thereby Luca has "reverted the reverts" in 19.11.6 [1].
> > > > > > > 
> > > > > > > But today we were made aware that still no OVS 2.13 builds against a
> > > > > > > DPDK that has those changes.
> > > > > > > Not 2.13.1 as we have it in Ubuntu nor (if it needs some OVS changes
> > > > > > > backported) the recent 2.13.3 does build.
> > > > > > > They still fail with the very same issue I reported [2] back then.
> > > > > > > 
> > > > > > > Unfortunately I have just released 19.11.7 so I can't revert them
> > > > > > > there - but OTOH reverting and counter reverting every other release
> > > > > > > seems wrong anyway.
> > > > > 
> > > > > It is wrong indeed, but the main question here is why these patches was
> > > > > backported to stable release in a first place?
> > > > > 
> > > > > Looking at these patches, they are not actual bug fixes but more like "nice to
> > > > > have" features that additionally breaks the way application links with DPDK.
> > > > > Stuff like that should not be acceptable to the stable release without a strong
> > > > > justification or, at least, testing with actual applications.
> > > 
> > > I agree, but TBH IIRC these changes were initially by OVS people :-)
> > > One could chase down the old talks between Luca and the requesters, but I don't
> > > think that gains us that much.
> > > 
> > > > > Since we already have a revert of revert, revert of revert of revert doesn't
> > > > > seem so bad.
> > > 
> > > As long as we don't extend this series, yeah
> > > 
> > > > > > > I wanted to ask if there is a set of patches that OVS would need to
> > > > > > > backport to 2.13.x to make this work?
> > > > > > > If they could be identified and prepared Distros could use them on
> > > > > > > 2.13.3 asap and 2.13.4 could officially release them for OVS later on.
> > > > > > > 
> > > > > > > But for that we'd need a hint which OVS changes that would need to be.
> > > > > > > All I know atm is from the testing reports on DPDK it seems that OVS
> > > > > > > 2.14.3 and 2.15 are happy with the new DPDK code.
> > > > > > > Do you have pointers on what 2.13.3 would need to get backported to
> > > > > > > work again in regard to this build issue.
> > > > > > 
> > > > > > You would need to use partial contents from patch :
> > > > > > http://patchwork.ozlabs.org/project/openvswitch/patch/1608142365-
> > > > > 26215
> > > > > > -1-git-send-email-ian.stokes@intel.com/
> > > > > > 
> > > > > > If you'd like me to send patches which would work with 2.13, 2.14, I'm
> > > > > > ok with that too.[keeping only those parts from patch which fixes the issue
> > > > > you see.] But we must ensure it doesn’t cause problems for OVS too.
> > > > > > Your thoughts Ilya ?
> > > > > 
> > > > > We had more fixes on top of this particular patch and I'd like to not cherry-
> > > > > pick and re-check all of this again.
> > > > 
> > > > I agree, we had more fixes on top of this. It would be risky to cherry-pick.
> > > > So it might be a better option to revert.
> > > 
> > > I agree, as far as I assessed the situation it would mean the revert
> > > of the following list.
> > > And since that is a lot of "reverts" in the string, to be clear it means that
> > > those original changes would not be present anymore in 19.11.x.
> > > 
> > > f49248a990 Revert "Revert "build/pkg-config: prevent overlinking""
> > > 39586a4cf0 Revert "Revert "build/pkg-config: improve static linking flags""
> > > 906e935a1f Revert "Revert "build/pkg-config: output drivers first for
> > > static build""
> > > deebf95239 Revert "Revert "build/pkg-config: move pkg-config file creation""
> > > a3bd9a34bf Revert "Revert "build: always link whole DPDK static libraries""
> > > d4bc124438 Revert "Revert "devtools: test static linkage with pkg-config""
> > > 
> > > But to avoid going back&forth I'd prefer to have a signed-off on that
> > > approach from:
> > > - Luca (for 19.11.6 which has added the changes)
> > > - Bruce (for being involved in the old&new case in general)
> > > - Thomas (for general master maintainer thoughts)
> > > 
> > 
> > If this is what is needed to ensure OVS can continue to use this release
> > series, then I am absolutely fine with it.
> 
> This was requested by OVS, so if they don't need it anymore it's fine
> by me as well

I am not sure to understand the full story,
but I am a bit worried that our release is dictated by
a single "user" (project using DPDK).

Please do you have links of discussion history?