[Piuparts-devel] [stretch] Status on upgrade paths

[Niels Thykier]

Andreas Beckmann:
> On 2017-04-12 13:26, Niels Thykier wrote:
>> I see you have been busy filing bugs for issues in the upgrade path.
>> Many thanks for that, it is greatly appreciated. :)
>>
>> Do you have an overview of how the upgrade paths are looking these days?
>>  Like how much have we tested vs. how much is missing?
> 
> I don't think I have much backlog left (usually taking logs from my 
> piuparts instance, we only recently started running more complicated 
> upgrade paths on piuparts.d.o).

Thanks, that is great to hear. :)

> But I don't always catch buggy uploads to sid that are unblocked+aged/2 :-)
> 

Are these bugs something we could (reasonably) stop via the Britney
integration for buster?  Admittedly, it would be "always" on, so we have
to be careful that we have few false positives on the RC bugginess of
the issue spotted by piuparts.

In particular, the the symlink_to_dir/dir_to_symlink migrations bugs
appears to be quite common, so I would be interested in getting them
stopped at Britney if possible.

> Concerning "kept back" packages:
> 
> https://piuparts.debian.org/jessie2stretch/packages_have_been_kept_back_issue.html
> https://piuparts.debian.org/jessie2stretch-rcmd/packages_have_been_kept_back_issue.html
> (these lists are too long, retesting them is queued, most should be
> fixed by libstdc++6 now having Breaks: libopenmpi1.6)
> * ignore packages no longer existing in stretch (version=None)
> * a nasty one is jakuk
                   ^^^^^

Is that the actual package name?  I cannot find any patches for it in
unstable.

> [...]
> * code-saturne-bin seems to get bitten by libopenmpi1.6 in stretch 
>   carrying a Conflicts: libopenmpi2 (which we have in stretch)
>   that is a bug in apt (but I don't think I even filed that one)
>   see #859986 against libopenmpi1.6 in jessie which must be fixed
>   in the upcoming point release (otherwise it's hopefully too late)
>   but a better solution would be welcome that does not require
>   fixes in stable
> 

Worst case, we will have to document it in the release notes until it is
fixed in jessie.

> [...]
> * gitlab is waiting for migration (hint outdated), sid version is
>   probably not worse than (uninstallable) stretch version
> 

Ok, I got a few remarks for it, so I will follow up with the maintainer.
 But I suspect you are right the that the sid version is better than the
stretch version atm.

> * metaphlan2-data and pv-grub-menu have downgraded bugs, OK for me
> 
> 
> Andreas
> 

Ok, thanks, :)
~Niels