Martin -

I'll let you know when take-3 is ready.

donald

On Wed, Nov 18, 2015 at 9:09 AM, Martin Winter <
[email protected]> wrote:

> On 18 Nov 2015, at 5:45, Donald Sharp wrote:
>
> I assume you mean --enable-multipath=0?  I have a fix on take-3:
>>
>
> Yes.
>
>
>> https://github.com/donaldsharp/quagga/commit/2810e353679b888036719cd5f21c7431c51ba305
>>
>> I'm in the middle of addressing CR comments for the Nexthop Tracking
>> Patch.... The changes there are going to ripple pretty significantly down
>> the line.  So expect another day or two or three before I can wrap up the
>> next iteration.  Plus I need to find the bug fixes for the issues reported
>> and add them to the branch as well.
>>
>
> I think I’ll suspend the testing now on take-2 and restart it on take-3
> It seems a lot of the failures are based on the same issues:
>         access-list crash  (breaks many of my BGP tests)
>         enable-multipath=0  (breaks about all IPv6 tests which verify with
> packet forwarding)
>         neighbor <if-name> interface  (breaks all ospfv2 virtual link
> tests)
>
> Not sure if you manage to get fixes (specially the last which I just
> reported) for
> all of these into take-3, but seems like enough changes go in there to just
> move testing over there. I’ll probably start the main testing on take-3 in
> 22 hrs
> (when I have time again)
>
> CI system test plan for your patches is now changed over to take-3
>  https://ci1.netdef.org/browse/QUAGGA-CUMULUS
>
> Currently running the first pass and should automatically kick in every
> time
> you submit more patches to it. Also added the OSPF-4.15 test to the short
> list
> of protocol checks which runs every time.
>
> Regards,
>    Martin Winter
>
>
>
> On Wed, Nov 18, 2015 at 8:11 AM, Martin Winter <
>> [email protected]> wrote:
>>
>> Was wrong here.
>>>
>>> The RIPNG issues are the multipath configure option as well.
>>>
>>> - Martin Winter
>>> [email protected]
>>>
>>>
>>> On 17 Nov 2015, at 7:28, Martin Winter wrote:
>>>
>>> Give me at least another day before sending out a take-3
>>>
>>>> Currently looking into the RIPNG failures (they seem to be something
>>>> else and not the multipath issue). Maybe we can get that fixed
>>>> as well in the same pass…
>>>>
>>>> - Martin
>>>>
>>>>
>>>> On 17 Nov 2015, at 7:16, Donald Sharp wrote:
>>>>
>>>> Yep.  I'm fixing this for take-3.  It's an easy fix.
>>>>
>>>>>
>>>>> donald
>>>>>
>>>>> On Tue, Nov 17, 2015 at 10:09 AM, Martin Winter <
>>>>> [email protected]> wrote:
>>>>>
>>>>> On 17 Nov 2015, at 6:59, Donald Sharp wrote:
>>>>>
>>>>>>
>>>>>> We build an array to handle the multipaths, when a 0 is specified it
>>>>>>
>>>>>>> builds
>>>>>>> an array of size 0 :)  Hence the problem.  This does need to be
>>>>>>> fixed.
>>>>>>>
>>>>>>>
>>>>>>> Ah.. so I assume you already know where to look for this one and I
>>>>>> can
>>>>>> move
>>>>>> on to the next issue?
>>>>>>
>>>>>> - Martin
>>>>>>
>>>>>>
>>>>>> On Tue, Nov 17, 2015 at 9:55 AM, Martin Winter <
>>>>>>
>>>>>> [email protected]> wrote:
>>>>>>>
>>>>>>> Donald,
>>>>>>>
>>>>>>>
>>>>>>>> I’m working on the next issue right now and leave you with the
>>>>>>>> crash.
>>>>>>>> I assumed you might have a faster idea where to find the problem
>>>>>>>> (and
>>>>>>>> fix)
>>>>>>>>
>>>>>>>> Currently looking at BGP IPv6 issues. It seems the
>>>>>>>> “--enable-multipath=0”
>>>>>>>> got broken (again). “--enable-multipath=64” seems to work, but with
>>>>>>>> a value of 0 (= unlimited), something breaks in getting IPv6 routes
>>>>>>>> installed.
>>>>>>>>
>>>>>>>> - Martin
>>>>>>>>
>>>>>>>> On 17 Nov 2015, at 6:51, Donald Sharp wrote:
>>>>>>>>
>>>>>>>> Martin -
>>>>>>>>
>>>>>>>>
>>>>>>>>> Before you narrow down a specific patch that might be causing the
>>>>>>>>>
>>>>>>>>> issue...
>>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> I've been contemplating working through our entire patch set and
>>>>>>>>> exposing
>>>>>>>>> it through the take-X branches.  I can't imagine we haven't already
>>>>>>>>> fixed
>>>>>>>>> the crash you are seeing( and I'm looking for that by hand now ).
>>>>>>>>>
>>>>>>>>> In addition I think you've provided enough of a clue for me to
>>>>>>>>> narrow
>>>>>>>>>
>>>>>>>>> down
>>>>>>>>>
>>>>>>>>
>>>>>>>> the problem to the actual bit of code and to find our fix.
>>>>>>>>
>>>>>>>>>
>>>>>>>>> donald
>>>>>>>>>
>>>>>>>>> On Tue, Nov 17, 2015 at 9:43 AM, Martin Winter <
>>>>>>>>> [email protected]> wrote:
>>>>>>>>>
>>>>>>>>> Han,
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>> On 17 Nov 2015, at 4:27, Han Coumans wrote:
>>>>>>>>>>
>>>>>>>>>> Dear Martin, Donald, Quagga devlist,
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> From CI's RIPng RFC test we learn that even without a new patch
>>>>>>>>>>> (from
>>>>>>>>>>>
>>>>>>>>>>> the
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>> Cumulus Mega patch set 'Take 2') for RIPng it's possible to have 10
>>>>>>>>
>>>>>>>>> new
>>>>>>>>>
>>>>>>>>> failures, making a total of 17 (one is an 'unpredict').
>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Keep in mind to look at the differences. The cumulus mege set is
>>>>>>>>>> based
>>>>>>>>>>
>>>>>>>>>> on
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>> top of the last round (accepted 4) - so the main thing is to compare
>>>>>>>>
>>>>>>>>> to
>>>>>>>>>
>>>>>>>>> this one.
>>>>>>>>>>
>>>>>>>>>> And to give kudos to Cumulus: There are actually quite a few past
>>>>>>>>>>
>>>>>>>>>> failures
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>> which show up as fixed.
>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>> For the failed ones, my experience is that a single issue (i.e. in
>>>>>>>>>>
>>>>>>>>>> zebra)
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>> might impact many tests. I’m still working through the failures and
>>>>>>>>
>>>>>>>>> will
>>>>>>>>>
>>>>>>>>> post whenever I find a specific one (and can explain the exact
>>>>>>>>>> issue or
>>>>>>>>>> commit or how to reproduce). But it will take some time with such
>>>>>>>>>> a
>>>>>>>>>>
>>>>>>>>>> large
>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>> set.
>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>> Regards,
>>>>>>>>>>
>>>>>>>>>> - Martin Winter
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> That will be an interesting fix.
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Thanks for your CI, Martin.
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> Regards,
>>>>>>>>>>> Han Coumans
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>
_______________________________________________
Quagga-dev mailing list
[email protected]
https://lists.quagga.net/mailman/listinfo/quagga-dev

Reply via email to