[ovs-dev] [PATCH ovn 0/2] binding: Fix incremental processing of child Port_Bindings.

[Dumitru Ceara]

On 1/19/21 10:27 AM, Numan Siddique wrote:
> 
> 
> On Tue, Jan 19, 2021 at 2:43 PM Dumitru Ceara <dceara at redhat.com 
> <mailto:dceara at redhat.com>> wrote:
> 
>     On 1/19/21 8:32 AM, Numan Siddique wrote:
>      >
>      >
>      > On Mon, Jan 18, 2021 at 10:20 PM Dumitru Ceara <dceara at redhat.com
>     <mailto:dceara at redhat.com>
>      > <mailto:dceara at redhat.com <mailto:dceara at redhat.com>>> wrote:
>      >
>      >     Dumitru Ceara (2):
>      >            binding: Fix container port removal from local bindings.
>      >            binding: Always delete child port bindings first.
>      >
>      >
>      > Thanks Dumitru for fixing these issues. I applied these patches to
>      > master and back ported to 20.12 and 20.09 branches.
>      >
> 
>     Thanks, Numan, for backporting the fixes.  However, on branch 20.09,
>     even if the patch applied cleanly the tests don't really execute
>     properly because of the missing test utilities
>     (check/fetch_column/etc.):
> 
>     testsuite.dir/at-groups/153/test-source: line 24: check: command not
>     found
>     testsuite.dir/at-groups/153/test-source: line 28: check: command not
>     found
>     testsuite.dir/at-groups/153/test-source: line 31: wait_row_count:
>     command not found
>     testsuite.dir/at-groups/153/test-source: line 32: fetch_column: command
>     not found
>     testsuite.dir/at-groups/153/test-source: line 33: wait_row_count:
>     command not found
>     testsuite.dir/at-groups/153/test-source: line 34: wait_row_count:
>     command not found
> 
> 
> All the tests for branch-20.09 passed on my local setup and on CI - 
> https://github.com/ovn-org/ovn/runs/1726114597 
> <https://github.com/ovn-org/ovn/runs/1726114597>

Yes, the test passes but it doesn't actually test for anything. :)

> 
> And I didn't notice the missing check.
> 
>     Shall we consider a branch-20.09 specific patch to fix the test?
> 
> 
> We may see similar issues for future backports too.
> 
> I think we can either backport the commit which added the "check" 
> library function or
> have a branch-20.09 patch which stubs check function so that we don't 
> see such issues ?

I think backporting the commit that adds "check" to the unit tests is 
probably the best way to go.  It's not exactly a feature and makes 
future backports easier.

Regards,
Dumitru