Skip to content

Standup Notes 2018 11 29

Erik Moeller edited this page Nov 30, 2018 · 1 revision

Participants (alphabetical): Conor, Emmanuel, Erik, Jaysinh, Jen, Kevin, Kushal, Mike, Mickael, Nina

Call-outs / topics for extended agenda

  • Release: Two new strings in the journalist interface -- require translations. Greek and Romanian almost ready. Kushal will file issue.

Conor

Yesterday: Reviewed & merged Ansible upgrade PR.

Today: Reviewed and merge 4.4.162 kernel config PR (submitted a month ago, just cleaning up). Today targeting CI docs so we can get in the GCE logic. Xenial upgrade tasks likely not till tomorrow. PSA, I'll be logging out a bit early today.

Blockers: None

Erik

Yesterday:

  • Hiring
  • Sprint planning & follow-up
  • UX research planning
  • Light spec/issues work
  • Qubes debugging & client testing

Today:

  • Hiring
  • Spec/issues work
  • Prototype QA

Blockers:

None

Jaysinh:

Yesterday:

  • Tried browsing the Kernel patching in Ubuntu.
  • Observed the grsecurity kernel configurations.

Today:

Blockers:

  • ansible-role-grsecurity requires grsecurity subscription (paid). -> Should not be required for wireless removal task

Jen

Yesterday:

  • Hiring stuff
  • Minor 0.11 release related tasks, e.g. finished testing mickael's ansible PR
  • Review/merged QA loader
  • Towards EOD started review gpg wrapper, will finish today

Today:

  • Finish review of Gpg wrapper in client
  • Wee bit more hiring stuff in PM
  • Standing by to do 0.11 release bugfixes as needed

Blockers:

None

Kevin

Yesterday: Bit of support work. 16.04 install on NUC hardware.

Today: Made edits to QA test plan; recorded thoughts on how Xenial upgrade could work. More time prepping for QA. Back to NUC.

Blockers: None

Kushal

Yesterday:

  • reviewed and merged "remove ini config file (#31)"
  • reviewed and merged "remove ini config file #188 on securedrop-client"
  • Updated forum for 2 new strings
  • Informed translators/team about the same.

Today:

  • will update the test plan from Kevin's input

Blockers: None

Mike

Yesterday: Nothing!

Today: winding down ASRE Ops on-boarding

Blockers: Qubes fires

Mickael

Yesterday:

Built RC debs for 0.11.0. As part of QA -- maintainers, please apply patch that I posted to internal tracker. Found server misconfig.

Today:

QA on 0.11.0

Blockers:

None

Nina

Yesterday:

Prototyping

Today:

UX meeting this morning

Am working on prototype to get it ready for QA.

Blockers:

None

Saptak

No updates

Extended discussion

Python packaging workflow -- ways to reduce complexity?

Jen: We have complicated workflow for incorporating Python wheels we build into debs. See diagram: https://github.com/freedomofpress/securedrop-debian-packaging#securedrop-debian-packaging

We should do one of two things: 1) think about ways to reduce complexity, 2) spend more time documenting process. Thoughts?

Kushal: Wrote a blog post explaining the whole process: https://kushaldas.in/posts/building-wheels-and-debian-packages-for-securedrop-on-qubes-os.html

We should think more about source tarballs we're using.

Jen: How could we do a better job verifying sources? Inspect the diffs?

Kushal: Don't have solid answer -- teams I'm working with manually inspect changes

Mickael: Let's carefully examine which dependencies are absolutely necessary. Looking at diffs for some libs is feasible, for others it may be more complicated. Would it make sense to have the wheels built at the time the packages are built? Having the wheels prebuilt does not offer an increment in assurances or security, as far as I can tell.

Kushal: We would lose deb reproducibility because wheels would be different.

Conor: As I understand it we don't have that as it is.

Mickael: wheel process itself is not reproducible. We are investing trust in person building the packages.

Kushal: That's why I asked around what other folks are doing. Almost all of them maintain their own repository like we do. Have a whitelist of packages.

Mickael: We may be conflating building and retrieval for wheels vs. build process for packages.

Kushal: Having dedicated machine in known state for building wheels and debian packages. Idea was to do this in future, as I understand.

Jen: Sounds like we agree we should have dedicated machine for the build process. What do we gain from building/pushing wheels separately?

Conor: dh-virtualenv project is most relevant for workstation. We have separate build process for SecureDrop servers, using Python/django code elsewhere. As we expand, perhaps in certain contexts, we ship application code in containers. In order to get app code to land where we need it in a way that's reliable and auditable, perhaps we can find ways to simplify.

What we've done in the past is to have dedicated hardware, disconnected unless needed.

Jen: Let's not forget we have potential uses for dh-virtualenv on server as well.

Near-term action: Architecture conversation about automated build processes.

Clone this wiki locally