mongodb xenial s390x packages are needed (blocks ceilometer)

fyi, there's a complete untested build now available here:
  lp:ubuntu-s390x-community/mongodb

er ppa:ubuntu-s390x-community/mongodb

We probably want to do something like this to enable the mongodb charm for z:
  https://code.launchpad.net/~dannf/charms/trusty/mongodb/arm64-default-to-ppa

If someone wants to work on enabling mongodb on s390x for zesty, see http://bugs.debian.org/848298

This bug was fixed in the package mongodb - 1:3.4.1-2

---------------
mongodb (1:3.4.1-2) experimental; urgency=medium

  * Disable hardware-assisted CRC32 on arm64 and s390x. This fixes FTBFS on
    these architectures.

 -- Apollon Oikonomopoulos <email address hidden> Fri, 23 Dec 2016 20:59:49 +0200

mongodb (1:3.4.1-1) experimental; urgency=medium

  * New upstream stable series
    + d/watch: look for 3.4 stable releases
    + Update upstream's signing key for 3.4
    + Drop fix-boost-1.60-build.patch; applied upstream
    + Refresh remaining patches
    + B-D on libboost-iostreams-dev
  * Upload to experimental
  * New patches:
    + Use std::regex instead of boost::regex in mongo shell (fixes FTBFS with
      Debian's boost and C++11)
  * Also build for s390x (Closes: #848298)
  * Drop i386 builds; i386 has long been deprecated upstream and support has
    been essentially removed in 3.4.
  * Use tcmalloc on ppc64el and arm64 as well
  * Do not ship mongosniff, as it is no longer built by the mongodb source
  * d/changelog: restore the 2.6.12-3 entry (Closes: #848305)

 -- Apollon Oikonomopoulos <email address hidden> Fri, 23 Dec 2016 14:57:21 +0200

Status changed to 'Confirmed' because the bug affects multiple users.

For Xenial (and Yakkety) the mongodb-server package needs to be taken from this ppa:
ppa:ubuntu-s390x-community/mongodb
see: https://launchpad.net/~ubuntu-s390x-community/+archive/ubuntu/mongodb

Given the major version difference to Xenial, I've marked that wont-fix in distro.

Hi,
the sync [1] of a mongodb set this to fix released quite a while ago, but in fact it fails to build on arm64 [2].
To make things less debuggable, the build issue seems to be a hang, killed by builders timeout and not a clear "error: foo is wrong".

I was "encouraged :-)" [3] for my s390x experience to look into this since s390x wanted the newer version of mongodb to get rid of the extra ppa [4] solution we have atm (see comment #8 in this bug).
But while I took a few tests with cross compiles and super slow arm emulations I couldn't get any closer to the issue. After all I'm not the arm expert, so maybe we should reconsider who owns the current task.

Now this issue is kind of back where it started. This is the bug that was referenced on IRC as "wanted s390x support" to begin with and holds said ppa solution for Xenial.
Fortunately Dannf is susbcribed on this bug and a much bigger arm expert which is what is really needed now to look into [2].

Furthermore to make things worse according to Steve langasek in bug 1679792 this now is a all-arch FTBFS for gcc-7 on top of the former issues.

Finally mongodb 3.4 has several more point releases (3.4.7 since last week), but that is a question to Debian if they are not updated for a reason.

IMHO this should be:
- tracked by JFH (for s390x needing the new version of mongodb)
- if possible debugged by Dannf (for arm expertise and the owner/driver of the former 3.2.4 ppa build)
- Both are also the former drivers/owners of this bug which just seems not resolved.

Since both are subscribed here already driving the old effort, maybe you just need this ping to realize it is not that much "fix released" as everybody thought (setting the task status now).

[1]: https://launchpad.net/ubuntu/+source/mongodb/1:3.4.1-3
[2]: https://launchpadlibrarian.net/311346682/buildlog_ubuntu-zesty-arm64.mongodb_1%3A3.4.1-3_CANCELLING.txt.gz
[3]: https://irclogs.ubuntu.com/2017/08/11/%23ubuntu-release.html#t18:15
[4]: https://launchpad.net/~ubuntu-s390x-community/+archive/ubuntu/mongodb

Hi,

Debian just uploaded 3.4.7 to experimental so I synced that. The new version fixes the gcc7 build issue.

The hang in the arm64 build happens during the build tests. We could skip the build tests on arm64 … if we think that the package will still work on arm64.

https://launchpad.net/ubuntu/+source/mongodb/1:3.4.7-1

On Mon, Aug 14, 2017 at 2:50 AM, ChristianEhrhardt
<email address hidden> wrote:
> Hi,
> the sync [1] of a mongodb set this to fix released quite a while ago, but in fact it fails to build on arm64 [2].
> To make things less debuggable, the build issue seems to be a hang, killed by builders timeout and not a clear "error: foo is wrong".
>
> I was "encouraged :-)" [3] for my s390x experience to look into this since s390x wanted the newer version of mongodb to get rid of the extra ppa [4] solution we have atm (see comment #8 in this bug).
> But while I took a few tests with cross compiles and super slow arm emulations I couldn't get any closer to the issue. After all I'm not the arm expert, so maybe we should reconsider who owns the current task.
>
> Now this issue is kind of back where it started. This is the bug that was referenced on IRC as "wanted s390x support" to begin with and holds said ppa solution for Xenial.
> Fortunately Dannf is susbcribed on this bug and a much bigger arm expert which is what is really needed now to look into [2].
>
> Furthermore to make things worse according to Steve langasek in bug
> 1679792 this now is a all-arch FTBFS for gcc-7 on top of the former
> issues.
>
> Finally mongodb 3.4 has several more point releases (3.4.7 since last
> week), but that is a question to Debian if they are not updated for a
> reason.

Good news is that mongodb 3.4.7 was uploaded to experimental
yesterday, and successfully builds for arm64 there. It also appears to
have a patch to fix compilation w/ gcc-7. Bad news is that it still
FTBFS on arm64 in artful:
  https://launchpad.net/~dannf/+archive/ubuntu/mongodb/+build/13246095

> IMHO this should be:
> - tracked by JFH (for s390x needing the new version of mongodb)
> - if possible debugged by Dannf (for arm expertise and the owner/driver of the former 3.2.4 ppa build)

I'll spend some time to see if I can understand why it builds in
Debian and not Ubuntu. But, TBH, I don't have any knowledge of mongodb
internals - I only sponsored the uploads to the mongo/s390x PPA - so I
am really just looking for a light switch in the dark :)

  -dann

dann, it looks like you might have missed my comment on this bug earlier today.

Also, could you go ahead and cancel your arm64 build that got stuck hours ago so the Launchpad builder can be available for other uses?

On Mon, Aug 14, 2017 at 9:11 PM, Jeremy Bicha <email address hidden> wrote:
> dann, it looks like you might have missed my comment on this bug earlier
> today.

Yep - waited until my build hung before sending, slow race w/ yours.

> Also, could you go ahead and cancel your arm64 build that got stuck
> hours ago so the Launchpad builder can be available for other uses?

done.

  -dann

The google-perftools version appears to be the significant difference between Debian and Ubuntu leading to the test hangs. I'm working on a debian merge of that package to identify the relevant delta.

This bug was fixed in the package google-perftools - 2.5-2.2ubuntu1

---------------
google-perftools (2.5-2.2ubuntu1) artful; urgency=medium

  * Merge from Debian. Remaining changes:
    - debian/control: Use multiline syntax for Build-* fields.
    - Use 3.0 (quilt) format instead of cdbs' deprecated simple-patchsys.
    - Rewrite packaging using debhelper sequencer.
    - Build using -fno-strict-aliasing.
    - Build using recent autotools.
    - Build for s390x.
    - debian/patches/Use-initial-exec-tls-for-libunwind-s-recursion-flag.patch:
      Use initial-exec tls for libunwind's recursion flag.
  * Enable testing for all architectures by default.
  * Build-depend on libunwind-dev instead of libunwind8-dev, which is
    now a transitional package.
  * Build-depend on libunwind-dev on architectures where available.
  * Disable libunwind on arm64 due to hangs. (LP: #1595242)
  * Disable -fno-exceptions to fix test hangs on arm64.
  * Ignore test results on arm64 and armhf.

 -- dann frazier <email address hidden> Fri, 18 Aug 2017 16:36:56 -0600

dann, thanks. Now all we need is an Archive Admin to allow mongodb/i386 to be removed.

Seems in artful

mongodb | 1:3.4.7-1 | artful/universe | source, amd64, arm64, ppc64el, s390x