Qt Contributors Summit 2019 Program

Back to Qt Contributors Summit 2019

Please add a longer session description with topic owner in the lower part of the page!

Tuesday, 2019-11-19

First day is reserved for topics that are of interest to the majority, and main location is the Assembly Hall. If you have something to present please coordinate beforehand on IRC or via mail.

Time	Assembly Hall
8:00 - 9:00	Registration
9:00 - 9:20	Introduction and sponsors
9:20 - 10:30	Keynote: Towards Qt 6 (Lars Knoll)
10:30 - 11:00	Coffee Break
11:00 - 11:30	QML Version 3
11:30 - 12:00	CMake Port
12:00 - 13:30	Lunch Break
13:30 - 14:00	Branch Policy for Qt 6
14:00 - 14:30	Qt Marketplace
14:30 - 15:00	KDE experience in attracting and nurturing contributors
15:00 - 15:30	Coffee Break
15:30 - 16:00	Qt 6 Graphics Overview
16:00 - 16:30	Qt for Python in Qt 6
16:30 - 17:30	Agenda Overview

Wednesday, 2019-11-20

Time	Assembly Hall	1.3.14 (Zoo, 16 people)	1.1.9 (Landsberger Allee, 21 people)	1.1.8 (Greifwalder Str, 17 people)
9:00 - 9:40		Qt Marketplace	QtQml	Remote display of Qt applications in Qt 6
9:50 - 10:30	API Review Process			Qt 6 Network Overview
10:30 - 10:50	Coffee Break
10:50 - 11:30	Refurbishing Qt Widgets internals	Clang-based cpp parser for lupdate	Code Review: Sharing the load	Available, hidden and missing gems on the way of using Qt on embedded devices
11:40 - 12:20			Improve the contributor experience of the Qt project	Qt 6 Changes / Migration
12:20 - 13:20	Lunch Break
13:20 - 14:00	C++17 language and std library features for Qt 6		Qt application lifecycle (startup, shutdown, sessions)	Releasing
14:10 - 14:50		QtGUI, RHI & 3D
14:50 - 15:10	Coffee Break
15:10 - 15:50	BoF: Fuzzing Qt	Branching policy	Rethinking serialization for Qt6	HighDpi
16:00 - 17:00	Plenary Session (afterwards: Free Beers at the Reception Room)

Thursday, 2019-11-21

Time	Assembly Hall	1.3.14 (Zoo, 16 people)	1.1.9 (Landsberger Allee, 21 people)	1.1.8 (Greifwalder Str, 17 people)
9:00 - 9:40	QtCore	Qt CMake Workshop		Fate of Qt Solutions?
9:50 - 10:30
10:30 - 10:50	Coffee Break
10:50 - 11:30	Platform-specific APIs in Qt 6	Qt CMake Workshop	Evolving the Qt Project Security Policy	Qt for Python and beyond
11:40 - 12:20			Qt WebEngine Release Management	Qt for Python Documentation
12:20 - 13:20	Lunch Break
13:20 - 14:00	moc and QMetaObject	Qt CMake Workshop	Future of QStyle for widgets and controls
14:10 - 14:50		Qt CMake Workshop	Qt Wayland Client and extensions	Qt Machine Learning and Math
15:00 - 16:00	Plenary Session

Sessions

QtCore

Thiago Macieira

Qt Contributor Summit 2019 - QtCore

Agenda Overview

The Qt Contributors Summit is mostly organized as an un-conference. In addition to the sessions already proposed here, we assume that participants will come up with topics that they'd like to discuss or work on. At the end of Day 1, anyone that has a session proposal will have a chance to pitch their session, and all participants can indicate which sessions they'd like to participate in. We'll then create the agenda for days 2 and 3, trying to match the sessions with the biggest interests to the largest rooms. The agenda will be updated here.

QML Version 3

Ulf Hermann

CMake Port

Alexandru Croitor

State of Qt6 port to CMake.

Branch Policy for Qt 6

Volker Hilsheimer

The current flow of changes in which fixes go into a stable branch, and are then up-merged gradually through several branches before they reach dev, can cause significant delays, results in a lot of changes in those stable branches, requires a lot of process work that provides very little enjoyment to the people responsible for it, and clutters the dev branch with merge commits. The approach has some benefits as long as it applies cleanly, i.e. it avoids duplicate commits, but with the dev branch having become Qt 6 with large scale refactoring and incompatible changes, it doesn't apply cleanly anymore anyway.

A simpler workflow where all changes go into the dev branch, and are then cherry-picked into those stable branches where we want them, perhaps supported by some automation, seems to be more common in software projects. It provides different trade-offs, which we should discuss.

KDE experience in attracting and nurturing contributors

David Edmundson

KDE has been a primarily volunteer driven project for its 20+ year lifespan. Attracting and keeping volunteers can be a difficult process, but there are many easy steps that can have a positive return. By sharing what we do successfully in KDE, we hope to kickstart a discussion of what steps we can emulate in Qt to improve the number of contributor contributions.

Qt Marketplace

Tino Pyssysalo and Marko Finnig

Qt Marketplace status, existing and planned features, launch schedule, content publishing process, and content usage.

Jira item: https://bugreports.qt.io/browse/QTPM-278

You may create bugs and suggestions in QtBug, but please link to the task above.

Publisher flow

• Publisher page
  • Rest APIs for contributing to Qt Marketplace would be more than welcome
• Repo/package publishing with the IFW
  • Very time consuming to create repos for each Qt platform and for each Qt patch version - especially if there is very few users for the patch version
    • Could the IFW package management changed so that packages would not be required for patch version, if nothing has changed
    • To create IFW packages for the marketplace may take several weeks
  • The number of binary installation is not expected to be large even in the future
    • Currently five extensions from four publishers
• Package manager
  • Would be needed for source building
  • Conan/vcpkg seem still the best options
    • Conan works with many build systems, including Yocto. Vcpkg woks only with CMake.
    • Conan index would help in managing repos. Needs to be checked, if the commercial license really needed
    • Conan UX considered significantly better
• Update frequency
  • For extensions in public/customer reports, the updates may be as frequent as needed
    • Should be limited though to avoid some cyberattacks
  • Qt Creator plugins distributed via the IFW would need to be validated in Qt Creator beta + final releases
    • Gives six weeks for the contributor to update plugins for the new Qt Creator version

User flow

• Building extensions
  • Qt Creator plugins
    • Packaging wizard exists, but not available in public yet
    • Qt Creator API freezing is not a feasible option
      • Plugin APIs need changes to make the maintenance work easier
      • What would be a minimal API set, which could be frozen and would bring value to plugin developers at the same time
      • Rather than freezing some of the Creator APIs, it could be more useful to provide Python APIs?
    • Having a Qt Creator environment available in the installer to help building plugins with the right compiler, headers etc. would be beneficial - this should already be the case
  • Building in the IDE or cloud
    • IDE most feasible option in the short term
    • Some contributors expect us to support IDE builds
• Binaries
  • Could we use bundles, containing plugins for several Qt Creator versions

Qt 6 Graphics Overview

Laszlo & co.

Let's have an overview of graphics related changes in Qt 6. The keynote will probably mention some of these, the goal in this session is to expand on them a bit. There can then be deeper individual discussions on specific topics during the next two days, if there is interest.

C++17 language and std library features for Qt 6

Ville Voutilainen

With Qt 6 we want to be able to use some C++ 17 language and std library features. Not all compilers and platforms we are going to care about by the time Qt 6 comes around will support evertyhing, so this needs a balanced discussion of up- and down-sides, leading to a pragmatic subset that we can rely on. See https://bugreports.qt.io/browse/QTBUG-77477 for details. See Cpp17FeaturesWeReallyWant for other ruminations.

Qt Contributor Summit 2019 - C++17 Features Notes

Code Review: Sharing the load

Eddy

Sometimes we wait and wait for anyone to review our changes. Sometimes we struggle to keep up with all the reviews we're asked to look at. How can we organise this so that no-one waits too long but all of us still have time to hack on code ? Time To Be Arranged.

API Review Process

Volker Hilsheimer

As experienced during Qt 5.13 and Qt 5.14 releases, some changes to APIs were getting feedback only very late in the release process, when the header diff was uploaded for sanity review. This seems a bit late. I'd like to see if we can find a more effective way of integrating API reviews into the general code review process. Perhaps changes that change public headers require a slightly different process than fixes and changes that touch only the implementation.

Session Notes

Evolving the Qt Project Security Policy

Volker Hilsheimer

During summer, the Qt Project Security Policy was moved from a wiki page into QUIP-15, and during that review process, some changes and additions were proposed to strengthen the project's capability to respond to security issues. Those changes were not taken into the QUIP as part of the move, since for the moment we only wanted to move the content. Suggestions included a clearer statement how security fixes are applied to LTS releases; the integration of CVE handling when disclosing vulnerabilities; the documentation of processes established by the Qt Company; and a general review of the way the "core team of developers" is organized, and operating.

Discussing these (and additional) proposals, and agreeing on what should become part of the policy (and thus the responsibility of the Qt project) is the purpose of this session. A proposal is available here: https://codereview.qt-project.org/c/meta/quips/+/278819

Qt for Python and beyond

Cristián Maureira-Fredes

After one year since the official release of Qt for Python we have been getting many new ideas for features to include in the next releases. Most of the features are explained in the latest blog post we wrote, but nevertheless we should try to build the next versions in favor of the Qt ecosystem, this means not only improving the Qt for Python project, but more like answering the question "How Qt for Python can improve the Qt project?". Please join us in this session to discuss how we can make Qt for Python a first-class citizen in the project by giving it more responsibility.

Improve the contributor experience of the Qt project

Cristián Maureira-Fredes

After the following steps:

• Creating Qt account,
• Creating Gerrit account,
• Agreeing with the CLA,
• Configure gerrit locally.

You are ready to start contributing.

If you think that it is too much, you should join this discussion. We aim to focus on having a welcoming and easy-to-do process for getting more people involved in contributing to Qt. Check the related task on JIRA

Session Notes

Platform-specific APIs in Qt 6

Tor Arne Vestbø, Johan Helsing, Paul Olav Tvete

There are currently multiple different ways of exposing platform-specific APIs in Qt (Qt Platform Headers, Qt Platform Native Interface, Qt Platform Support, and Qt Foo Extras among others). Qt 6 is a good time to take a look at how this can be improved and coordinated better.

Refurbishing Qt Widgets internals

Tor Arne Vestbø, Johan Helsing, Paul Olav Tvete

This is a session to discuss changes to the underlying architecture in Qt Widgets, such as the backing store and parent/child hierarchy.

Qt Contributor Summit 2019 - Refurbishing Qt Widgets internals

Future of QStyle for widgets and controls

Shawn Rutledge, Richard Gustavsen, Uwe Rathman

This is a session to discuss a potential architecture for styles that can be used to render both widgets and Qt Quick Controls, making the scene graph available for use in styles, perhaps using techniques similar to QSkinny, perhaps enabling Qt widgets and controls to be backed by native platform widgets (at least on some platforms where it's most necessary), etc.

Qt Wayland Client and extensions

Johan Helsing, Paul Olav Tvete

What existing extensions should we support, and should we propose any new extensions for wayland-protocols? What functionality is missing from Qt Wayland Client, and how can we improve cooperation with KDE and other stakeholders? (note that there is a separate session about the general approach to platform-specific APIs in future Qt versions.)

Fate of Qt Solutions

Friedemann Kleint

There are some gems in the current solutions (Singleton, property editor, etc), should they live on in Qt 6 and how?

Clang-based cpp parser for lupdate

Lucie Gerard

Introduction to the new cpp parser for lupdate, based on clang-tooling.

Topics:

• How to build a clang-tool?
• What are the requirement to use the new cpp parser?
• Why a new cpp parser (what is gained, what is lost)?
• How can cmake facilitate the use of the new clang-based parser?

moc and QMetaObject

Olivier Goffart

• What future do the Qt project wants for moc?
• Many changes are talked about for QMetaObject in Qt6: automatic QMetaType registration, hash-table based lookups, merging with the property cache, ...
• moc currently hasn't still fully be ported to support C++11 yet. Features such as raw literals, trailing returns and other will confuse moc. Should moc be updated for these features, or:
• Should we make a clang port of moc. clang tools are already used by qt creator/qdoc, maybe lupdate, so this wouldn't be a new dependency, but it would become a mandatory dependency. This would allow moc to be much better at handling "advanced" C++ features. But clang would probably be much slower at parsing than moc currently is.
• While moc has to stay to keep source compatibility with Qt5, should we allow users not to need to rely on moc using something similar to https://github.com/woboq/verdigris , or even encourage it? Recent change with QML3 which will force the use of a moc-json generated file goes against it.

Fuzzing Qt

Robert Löhning

Fuzzing could already find a couple of crashes in Qt, including security related. Let's have a birds of a feather session on how to make the most of it.

• How to try it locally?
• Which code needs fuzz testing the most?
  I think of functions that handle input from possibly untrusted sources. Please propose such or let me know about other criteria.
• What's missing to test Qt in oss-fuzz?

Qt CMake Workshop

Alexandru Croitor

Help out with building Qt6 with CMake. Provide feedback on things that are important to you.

Qt Machine Learning and Math

Cristián Maureira-Fredes

What C++ and more precisely, Qt, can provide to the Machine Learning community? What about simple Math? Python has shown us how a simple API can transform a whole language, just by providing the proper tools. The C++ performance could be a game-changer regarding Data Science, and it has been used in frameworks like PyTorch.

Qt WebEngine Release Management

Allan Sandfeld Jensen

How should WebEngine be release around Qt 6.0 and in Qt 6 times. Should WebEngine have a Chromium update for 5.15, a sort of 5.16. Should WebEngine support building with both Qt 5 and Qt 6 for a time?

Should Qt WebEngine decouple and how? It could soft follow the Qt releases but not maintain older versions and instead support building with older Qt. This requires changes to build system and packaging. Can we get rid of all private Qt api usage.

Qt for Python Documentation

Cristián Maureira-Fredes

Currently, there are many issues with the documentation of the project: C++ snippets, Toolchain, Not optimal structure, unclear flow, mix of tutorials, missing examples, etc. The idea of this session is to define both the design, and content of the Qt for Python documentation for future releases.

Rethinking serialization for Qt6

Arnaud Clère

Serialization is an old problem, still, we keep writing code to serialize C++ data in specific ways again and again. With Qt5 for instance, you may have to code: QDebug << to debug it, QDataStream << and >> to marshal it to another Qt application, use QSettings to make it persistent, QJson* or QXml* to convey it on the web, QCbor* for the IoT, and QAbstractModelItem for the DB/GUI. Even though such code needs to be customized here and there, it is mostly boilerplate code. So, can we make this simpler for Qt6?

Indeed, I will present a solution that enables to read/write C++ data from/to any of those APIs by defining a single function which can be easily customized to specific needs. Its runtime overhead being almost negligible, I will go on talking about many data formats from QDataStream to XML since that is where the actual performance/safety/interoperability tradeoffs are made. That should trigger an interesting discussion on the tradeoffs made by the only broadly implemented serialization for Qt types: QDataStream.

Finally, I hope to trigger enough interest from the community to review and polish this proposal, and to enable more serialization choices for most Qt6 types.

Available, hidden and missing gems on the way of using Qt on embedded devices

Vladimir Minenko

In this talk I would like to review how Qt as whole helps in making embedded devices. There is a variety of offerings, solutions and even real Qt modules dedicated to or just relevant for embedded devices for one or another use case. I think, the total is currently almost unbeatable and at the same time, it is very hard to find. What is that total and how can we make it more accessible? Should we? Is this something contributors should care? Is there an actual interest in the field? I wish to find out. Do you?

The slides are available here. The link expires after Jan 20, 2020.

Notes:

• There is general agreement that there is a need in Qt for a hand-on guidance on an application architecture applicable on embedded platforms
• Current examples and guides in the doc do not tell people how to create “serious” app right way. Most of them are focusing on a given Qt feature and some may even be a kind of misleading, e.g. https://doc-snapshots.qt.io/qt5-5.14/qtdoc-tutorials-alarms-example.html named as “Getting Started Programming with Qt Quick” tells people to write an app completely in JavaScript and even provides a simple model for a calendar written only in QML
• It was seen as doubtful if there should really be an new “application platform”, since this has a danger to become too specific and limit the flexibility Qt provides
• A form of “best practice” would be a better way to go instead of an own complete application architecture. The latter seems to be too sophisticated at the begining
• Addressing and reflecting this in the docs is mandatory, still there should be a reference implementation along with some dedicated examples
• Quite some embedded projects still do ONE app. Further splitting it in more sub-apps might be seen as an overkill. This also means that any new works on an application architecture should keep considering this as a valid case along multi-application and multi-process cases
• It still was not clear if such a work is of interest for a broad community or if it is for The Qt Company as the commercial entity behind Qt. This is due to the fact that Qt for Device Creation and other related projects, like Qt Automotive Suite are seen as commercial programs with a limited involvement from the community

Next steps:

• have a review discussion with stakeholders in The Qt Company
• Define the scope for the doc changes and an outline for the “best practice” guide and app

Releasing

Jani Heikkinen

Qt Contributor Summit 2019 - Releasing Notes

Remote display of Qt applications in Qt 6

Jan Arne Petersen, Christoph Sterz

Discussion about API changes which makes it easier to provide remote display of Qt applications. Besides the transfer of the images/textures also input handling (multi-seat?) is relevant.

Qt 6 Network Overview

Timur Pocheptsov / Mårten Nordheim

Let's have an overview of network related changes in Qt 6. The idea is to iterate through the existing list and verify nothing is missing and/or should be adapted.