Summer coding ideas for 2011

From FedoraProject

(Difference between revisions)
Jump to: navigation, search
(Per offlist mail with Marc, adding entry)
(New proposal: Unmasking packages (consider it!))
Line 218: Line 218:
  
 
''Mentor(s):'' Marc Schoenefeld , mschoene@redhat.com
 
''Mentor(s):'' Marc Schoenefeld , mschoene@redhat.com
 +
 +
 +
== Masking packages ==
 +
Mixing stable and unstable packages (ie from different repositories) is quite tiresome and hardly manageable for more than a few packets. What is missing is something like the Gentoo approach where you can unmask (or mask) certain unstable packages so that they will be pulled and installed into your otherwise stable-only environment. The user has to resolve any conflicts that may arise.
 +
 +
Apart from the luxury of more bleeding-edge packages available, this could also really get more people into testing of unstable packages. With changed policies regarding only stable updates / backports to a stable Fedora release, this would make Fedora still attractive for users that like to work with latest packages but refrain from rawhide for whatever reasons.

Revision as of 10:48, 4 March 2011

Other ideas that were incomplete or worth considering are found at Summer coding ideas for 2009.

Find an idea you like? Want to propose your own? See our Getting Started Guide:

http://groups.google.com/group/redhat-summer/web/gsoc-getting-started

Ideas for the JBoss community can be found here. We will be participating with JBoss this summer.

Contents

Turn smock into a real program

Status:

Summary of idea: smock is a hack that Dan and I wrote a few years ago that lets you build a chain of RPMs in mock. It works out all the dependencies and builds the packages in the right order, using the result from an earlier build to satisfy dependencies in a later build, and being almost completely automated. What really needs to happen is that mock is extended to support this as a native feature.

Contacts:

Mentor(s):

Notes:

Port Infrastructure TurboGears apps to TG2

Status: Proposed

Summary of idea: Several Fedora Infrastructure applications are written in TurboGears 1.0, and for longevity need to be ported to TG2. These include bodhi, mirrormanager, packagedb, elections, fas, smolt.

Contacts: Matt Domsch, Toshio Kuratomi

Mentor(s): mdomsch


MirrorManager enhancements

Status: Proposed

Summary of idea: MirrorManager provides the mirror lists to all Fedora systems. In addition to porting to TG2 (see above), several enhancements would be welcome:

  • Simplify creation of new MirrorManager instances (non-Fedora users, such as CentOS)
  • Simplify selection of mirrors within Cloud Providers on granularity other than netblocks and ASNs
  • Other items on the TODO list

Contacts: Matt Domsch

Mentor(s): mdomsch

mw enhancements

Status: Proposed

Summary of idea: [ mw] is a command-line program that pretends to be a version control system. It should eventually help people who are console nuts (like me) edit MediaWiki-based wikis without having to shout at their web browser.

Contacts: Ian Weller

Mentor(s): Ian Weller

Notes: This is also a proposed hackfest at FUDCon Tempe but I doubt we'll get as much as I'd like done there.

NIS support in SSSD

  • Status: Proposed
  • Summary of idea: Currently, the System Security Services Daemon supports only LDAP for network user identity. Support for NIS identities has been requested several times by end-users.
  • Contacts: Stephen Gallagher
  • Mentor(s): Stephen Gallagher

SUDO support in SSSD

  • Status: Proposed
  • Summary of idea: Sudo 1.8.0 will support a plugin interface for sudo authorization decisions. It would be excellent for SSSD to provide such a plugin to provide cached access to sudo information stored in the sudo LDAP schema. This would make it easier to maintain centralized sudo rules that also function while offline.
  • Contacts: Stephen Gallagher
  • Mentor(s): Stephen Gallagher
  • Notes: http://www.sudo.ws/sudo_plugin.man.html

Maven and PackageKit integration

Status: Proposed

Summary of idea: Currently when building with maven locally, all spec files need to specify build dependencies manually. Java libraries and maven plugins should have something like "Provide(groupId:artifactId:version:format)" so that "yum install org.apache.velocity:velocity" will correctly find and install velocity package. Second part of this project can be creating maven plugin asking you to install via PackageKit the package that provides the dependency.

Contacts: Stanislav Ochotnicky or Alexander Kurtakov

Mentor(s):

Notes: Basic knowledge of maven and rpm is required.

KDE Plasma and PackageKit integration

Status: Proposed

Summary of idea: Add automatically-generated rpm dependencies for kde/plasma related services (see also blog entry on the topic), and add packagekit hooks to use them (one concrete example would be for plasma dataengines (see also blog entry on the topic).

Contacts: Rex Dieter

Mentor(s):

Notes: Intermediate knowledge of rpm and/or kde plasma coding


Avahi yum repositories

Status: Proposed

Summary of idea: Add a yum plugin that will use avahi to find local mirrors of canonical repositories, and tools/scripts for local http/nfs servers to advertise what they have. It would also be nice to include proper nfs repository support in yum or a plugin, so the advertised repositories can be on NFS filesystems that are not already mounted locally on clients. The use on the clients should be transparent without any local yum repo config changes except perhaps for installing/enabling the plugin. Later anaconda could also be taught to use avahi to find repositories for installation.

Contacts: Roland McGrath

Mentor(s):

Notes: Divisible into various separate pieces that are worthwhile on their own.


Fill in some of the Missing nVidia and Radeon API pieces

Status: Proposed

Summary of idea: Add some of the missing 3d APIs for at least one of the major card types, considerably helping out the beleagured Radeon and Nvidia reverse-engineering efforts. There is enough currently (F14) missing that most 3d games do not function correctly, many of the underpinnings of FireFox 4 acceleration are blacklisted under Linux, and many other visual and other components cannot make correct use of the gpu(s) under Linux. Progress is abysmal for many reasons.

Contacts:

Mentor(s):

Notes: Requires good integration with the existing Radeon and nVidia teams and their testsuites


Fedora Medical

Status: In progress.

Summary of idea: Here, we are looking for a couple students who have some experience in RPM packaging, python, and bash. This would be a good opportunity to learn in depth packaging and fedora contributor ecosystem.

This is a work in progress and details can be found here: http://fedoraproject.org/wiki/SIGs/FedoraMedical

We are looking forward to do mainly packaging and getting them published to fedora repo. However, we will also be doing some tooling and associated works. So, python and bash will be required.

Understanding fedora package maintainer guideline is required. Having existing packages in fedora will be a plus. Also, the student should be interested in maintaining some of those packages after SOC.

Contacts: Susmit Shannigrahi <susmit at fedoraproject.org>

Mentor(s): Susmit Shannigrahi, (anyone want to co-mentor?)

Notes:

Improve critical dependency management in Yum

Status: Proposed (please review)

Summary of idea: The idea (roughly) is that Yum could detect when a package update is going to break some important functionality of the system because is a dependency of another package that won't be updated (for example: Yum is not detecting that when a kernel is updated, because the previous kernel is not deleted, and the dependency isn't broken). This happens once in a while with kernel and third party packages that provide support for devices that aren't supported by Fedora directly because of licensing restrictions: ie. Fedora updates the kernel, and RPM Fusion packages related to Broadcom wireless usually need 2-4 days to be updated, resulting in a non working wireless.

Contacts:

Mentor(s):

Notes: I don't feel qualified to be mentor.

Create a spin to aid students in education and to help them understand Linux and Fedora better

Status: Proposed

Summary of Idea: The idea is to create a spin which would be more student friendly. It should have some graph plotting tools, some circuit simulator, maybe an equation solving application, IDE for some basic coding and a few compilers which are taught in colleges and other relevant stuff. It should consist of extensive documentation to help students use these tools and Linux in general. A few interesting games and music player would be a plus.

Additionally, if possible, one can imagine to throw up a screen during install which would help in picking the age group of the student. The applications can be installed accordingly. For example, I would like to have a scrabble game for a small kid then a circuit simulator.

Contacts:

Mentor(s): jreznik and rrix mentored a spin project for Fedora Summer Coding last year. Maybe they could take this.

Create a Fedora Events System

Status: Proposed

Summary of Idea: The idea is to create a website which would allow displaying events, divided by region, list attendance, topics discussed and other points. For reference, see Ubuntu Loco Events.

Contacts:

Mentor(s): Nushio develops webapps as a dayjob and wouldn't mind making it a summer job too.

Smarter Yum Metadata

Status: Proposed

Summary of Idea: Share yum metadata among users of a system and also fetch them faster instead of downloading the whole chunk each time.

A git-based metadata store was proposed but objected to by yum maintainers. There could be other ways of doing this, as suggested in the bug report or by going ahead with the git idea -- I believe it can be done.

Contacts:

Mentor(s):

Integrating Fuzzing-based instruments in quality assurance tools

Summary of idea: Generate fuzzing testcases from the existing "make test", which is generally only as good as the package maintainer placed effort into this Extending the range of testcases in a generic way contributes to an enhanced scope of quality and security testing of software. Key implementation technique is providing mock "fuzzing" devices and transparent in-process instrumentation (fuzzing loops) the turn-around time and bug discovery rate will improve. Transparent integration in build tools (like rpm-build macros) lower the entry hurdle for developers and maintainers.

Key components:

-Fuzzing devices: By this we mean user space devices that manipulate given original data based on a given fuzzing strategy, the strategy chosen depends on the nature of the file type, implemented by plugins (for a start provide comprehensive plugin source for image types, gif png, multimedia, midi).

-Package QA process: We think that a prototypical integration in rpm-build (macro, pre-waive certain results ) will help (Fedora) developers package maintainers to transparently gain from these functionality during package build.

Target: It is expected that all posix variants can benefit from wrapping fuzzing algorithms behind devices, however due to the rpm-build macro support, we expect most benefit for rpm-based distros.


Contacts: Marc Schoenefeld , mschoene@redhat.com

Mentor(s): Marc Schoenefeld , mschoene@redhat.com


Masking packages

Mixing stable and unstable packages (ie from different repositories) is quite tiresome and hardly manageable for more than a few packets. What is missing is something like the Gentoo approach where you can unmask (or mask) certain unstable packages so that they will be pulled and installed into your otherwise stable-only environment. The user has to resolve any conflicts that may arise.

Apart from the luxury of more bleeding-edge packages available, this could also really get more people into testing of unstable packages. With changed policies regarding only stable updates / backports to a stable Fedora release, this would make Fedora still attractive for users that like to work with latest packages but refrain from rawhide for whatever reasons.