Re: [PATCH 1/2] vm_smoke_test: Drop wheezy from test

[Henning Schild <henning.schild@siemens.com>]

Am Mon, 14 May 2018 19:15:45 +0300
schrieb Alexander Smirnov <asmirnov@ilbers.de>:

> On 05/14/2018 07:01 PM, Henning Schild wrote:
> > Am Mon, 14 May 2018 16:49:14 +0300
> > schrieb Alexander Smirnov <asmirnov@ilbers.de>:
> >   
> >> On 05/11/2018 08:54 PM, Alexander Smirnov wrote:  
> >>> Wheezy requires old e2fsprogs (1.42-) installed on the host,
> >>> otherwise wheezy kernel won't be able to mount root partition.
> >>> After each build host upgrade this becomes more complicated to
> >>> maintain outdated tool. So it's time for wheezy to slightly go...
> >>>      
> >>
> >> If there are no objections, I'd like to apply this to have CI build
> >> green. Then I'll be able to eventually release master :-)  
> > 
> > I have no objections to dropping wheezy support. But if doing so the
> > real change is not removing it from the CI scripts and keeping it
> > around.
> > There are configs that need to be removed, docs that need to be
> > updated. Just "git grep wheezy" and you will find many more
> > referrence than just CI. We do not have a support matrix, but as
> > long as we have code hanging around, that suggests that a certain
> > distro is supported ... it should be working.
> >   
> 
> For sure, I understand, that there is multiconfig for wheezy that
> should also go.
> 
> > Dropping wheezy is a topic that probably deserves its own thread.  
> 
> I already planned to discuss wheezy's value after the release (this 
> week). But there are two points:
> 
> 1. There is an issue with ext4 image creation after switching to wic, 
> dmesg in QEMU sometimes shows the following error:
> 
> 8<--
>    EXT4-fs (mmcblk0): bad geometry: block count 247892 exceeds size
> of device (247808 blocks)
> 8<--
> 
> This problem is reproduced on wheezy image only. So I want to debug
> it first before completely dropping wheezy, to prevent possible error
> leaking.

The wic series does not even support building weezy images so i fail to
see the connection. One thing that comes to mind is that the builder
used to need e2fsprogs/jessie-backports, if you removed that you might
have your explanation.

> 2. I want to have green CI to perform release.
> 
> So I propose drop wheezy from CI as the first step to unblock the
> process.

Sounds like the VW approach ;). if under_test then pretend
But if you are planning to phase out weezie that might be an OK way to
do it.

Henning
 
> Alex