Potential Projects

From CDOT Wiki
Revision as of 16:30, 18 September 2008 by Cbishop2 (talk | contribs) (Research PGO-related bugs)
Jump to: navigation, search

Contents

Introduction

This is a list of potential projects that need people.

Students: If you'd like to work on one of these, move the chosen project to the Project List and create a page for your work based on the Sample Project template.

Note: OSD/DPS students are encouraged to select a Mozilla or OOo project. LUX students are encouraged to select a Fedora project. If you are interested in a project from the other community, please discuss it with your professor.

Open Source Community Members: We welcome your recommendations for potential projects. Please create an account on this Wiki and create a description for your proposed project below. Please list your contact info (just an IRC or FAS2 name is OK) as well as links to any related web pages as Resources for the proposed project. (Questions? Ask Chris Tyler or David Humphrey).

Potential Fedora Projects

Sample Project

This is a sample project stub. You can use the template for Sample Project in order to create a project page for one of the stubs below. This is how you 'sign-up' for a project.

NOTE: if someone has already created the project page, speak to this person and see if you can join them. If so, simply add your name to the Project Leader(s) page. Otherwise, you can become a contributor later.

Ideas Pool

Lots of good ideas are listed in Summer Coding Ideas (Fedora Wiki). These projects were originally proposed for Google Summer of Code projects; if you want to use one of them, please discuss it with Chris Tyler before selecting it, because the scope and mentorship model for GSOC is different from that of the LUX projects.

Free-open font packaging

The gratis release of a collection of web fonts in 1996 had a devastating effect on many font projects, drainning the font market in western countries and limiting the use of other typefaces to typography niches. However, this program has been discontinued and the gratis versions of those fonts are no longer updated. Since Unicode.org's codification pace of human scripts has not abated, and the font formats have evolved (with the OpenType specifications), this font set is increasingly obsolete. Also, many scripts were never covered, leaving entire world regions without mass access to a way to write their language in the digital age.

An operating system of browser that relies on the 1996 font set for its text rendering is not really free. And text is still our main information media.

Faced with this problem many individuals and organisations have started creating and publishing new fonts, but they lack the clout of multinationals to get successfully distributed, and are often poorly structured. The aim of this project would be to identify as many of those free fonts as possible and package them in Fedora. The result would serve as a core component of Fedora's art spin, and as basis for a richer international experience in Fedora (and derivatives such as OLPC)

This project is mainly targeted at free software (in this case fonts) distribution dynamics. You'll learn some rpm packaging skills, and be exposed to many different font projects worldwide, some big, some small, some efficient, some poorly run.

A project second stage would be to capitalize on the experience acquired and publish a set of distribution best practices for font authors, including recommended release composition, generic makefile, etc.

The Fedora Fonts special interest group will provide mentoring, technical support, and documentation. An initial list of fonts to package has already been published.

Contact: Nicolas Mailhot <nim AT fedoraproject.org>

Corresponding Source Web App

Finish design and implement web application that provides downloadable SRPMS for any package+tag in the Fedora Package Source Code Control system. While we provide SRPMS for all packages at release, the updates and rawhide trees churn their packages more rapidly, and will remove the koji-built SRPMS when the binary packages are removed. This would allow people to request source corresponding to the packages they have on ISO media or otherwise.

Resources: MattDomsch, http://git.fedorahosted.org/git/?p=correspondingsource.git;a=blob;f=DESIGN;hb=HEAD

Personal Koji Repositories

Koji Personal Repos (KoPeR) is a concept to allow Fedora maintainers to create personal repos of packages to build against and share with others.

Resources: http://fedoraproject.org/wiki/JesseKeating/KojiPersonalRepos

Contact: Jesse Keating

Port Wubi to Fedora

If upstream is willing to integrate it, port wubi-installer and make it more cross distribution. Otherwise, provide a similar implementation for Fedora.

http://wubi-installer.org/

Delta RPM support in Koji build system

* https://fedorahosted.org/bodhi/ticket/160


Benchmark Nightlife Scalability Issues

Using Puppet, or some other management tool, create a grid of a dozen machines, then turn that into 12,000 machine. Address the network characteristics - how much bandwidth do you need on different components, the communication mechanisms - does UDP work at such a scale, the grid topology - what happens when two nodes are behind firewalls.

Resource: Nightlife wiki page, Matthew Farrellee

Add IPv6 Support to Condor

Sooner or later we're going to have to live with IPv6. Many applications are very tied to IPv4 and that includes Condor. Understand what IP addresses really mean to Condor, like many existing systems, and what the implications of an IPv6 conversion means and how it can be done. There are many resources about IPv6 interfaces, but not many good transition guides, and fewer that address what the change can really mean to an application - Condor takes addresses and passes them as ascii text between components.

Resource: Nightlife wiki page, Matthew Farrellee

Update Condor Build System

At the foundation of many good applications is its build infrastructure. The latest generation of tools includes GNU make with a very powerful interpreted language and ever increasing functionality in autotools. Condor has a build system from a few generations back. It uses imake, originally designed for building X. While new systems are more powerful and provide enhanced functionality, there isn't always a clear mapping to features of older systems. Don't only find a way to use newer tools to build Condor, but find a way to use the power of current generation tools in ways that aren't just a straight mapping of need satisfied by older systems.

Resource: Nightlife wiki page, Matthew Farrellee

Create Test Suite for Condor

Along with a good build infrastructure, a strong test framework and methodology is key to survival of modern software. Condor is a system that is shipped on, currently, 14 different platforms - OS + Arch combinations. New features need to be verified to work on all those platforms and shown to cause no regressions in other features. Condor has a large test suite, but no suite covers everything or in sufficient depth. Pick a few features, say the VM Universe or Concurrency Limits, and demonstrate full understanding of them by testing all of the corner cases. This is far from a simple task, and requires more than a cursory understand of the system to properly complete. Understanding the complex interactions of distributed components, common patterns exist across many distribute systems, as well as intimate interactions with the OS is key.

Resource: Nightlife wiki page, Matthew Farrellee


Potential Fedora+Mozilla Projects

Sample Project

This is a sample project stub. You can use the template for Sample Project in order to create a project page for one of the stubs below. This is how you 'sign-up' for a project.

NOTE: if someone has already created the project page, speak to this person and see if you can join them. If so, simply add your name to the Project Leader(s) page. Otherwise, you can become a contributor later.


Create a MDRK Spin

The Mozilla Developer Resource Kit is a set of tools, code, and documentation intended to make it easy for new Mozilla developers to get up to speed. Package the MDRK components (including the software tools, a Moz source tree, and documentation) as RPM packages and then create a Fedora "spin" (Live + Installable DVD) of these packages.

Edit the spin image (from Revisor/Live CD Tools) so that, in addition to being a bootable/installable disc, the image can be run in a VM under Windows/Mac OSX. The disc image must be edited to include Windows/Mac OSX versions of the tools, so that if the disc is inserted into a running Windows or Mac OSX system, the appropriate version can be installed.

Resources: humph, ctyler, http://142.204.133.123/mxr, http://zenit.senecac.on.ca/wiki/dxr

Potential Mozilla Projects

Sample Project

This is a sample project stub. You can use the template for Sample Project in order to create a project page for one of the stubs below. This is how you 'sign-up' for a project.

NOTE: if someone has already created the project page, speak to this person and see if you can join them. If so, simply add your name to the Project Leader(s) page. Otherwise, you can become a contributor later.

Thunderbird SMTP Auto-Sensing

Modify Thunderbird so that it uses the correct SMTP server for your current network and IP. In other words, if you are at home, use your home ISP's SMTP, but if you use a school wireless network, switch to the school's SMTP. These various SMTP configurations should be manually controlled by the user, that is, you don't have to try and figure out which SMTP to use. Rather, it should be configurable in an options dialog.

References: #maildev

Firebug "linting" for portability problems

Lots of web developers use Firebug and Firefox for building their applications, but we want those apps to work well in other browsers as well. If Firebug knew about JS or CSS patterns that could cause problems in other browsers, it would make it much easier to have those applications work in all browsers.

Resources: mfinkle, robcee

XULRunner Application Packaging

Help to develop an automated packaging system for XULRunner applications.

Resources: plasticmillion, mfinkle, #mozpad, #prism

'Compact databases' feature for Mozilla calendaring applications (Lightning, Sunbird)

Add the possibility to compact the database for local storage calendars based on SQLite using the VACUUM command. At the moment the local database for events and tasks will not get smaller even if you delete your tasks and events. This has serious performance implications for people, who work a lot with their calendar.

See also Bug 352976

Resources: ctalbert, daniel in #calendar on irc.mozilla.org

Thunderbird Draft Scheduler Extension

Write an extension for Thunderbird so that the user has the ability to mark a draft email for sending later. This is not the same as having to save emails in your DRAFT box because users would have to remember to go back in and click Send later. What is different here is that users could set the date and time for each email to be sent. This feature would be helpful to those who work late into the night but prefer not to send emails at such a late hour. This feature would also benefit support staff, e.g., administrative assistants, who constantly have to send out regular email reminders, e.g., for meetings, timesheets, etc.

Resources: Vivian Ngo

Spellcheck Extension for Arbitrary Web Pages

Write an extension to leverage the existing spellcheck code in Mozilla and add the ability to highlight spelling mistakes for a given web page (i.e., vs. a textbox).

Modify Firefox to handle files downloaded to Temp more appropriately

Often files downloaded by the browser are put in a temporary folder that is emptied on close. Users (and especially novice users) should be protected from inadvertent data loss as a result of important files being saved to this temp folder. This project will add fixes to the browser so that users are protected. Ideas include:

  • Make Firefox aware of common productivity file types (e.g., .doc, .pdf) and whitelist these types so they aren't deleted
  • Have office type files download automatically to a documents folder instead of temp
  • Firefox could check the timestamp of such files and delete them only if the timestamp is the same as when it was originally saved by the browser (i.e., has not been edited)
  • Warn the user on closing the browser that files are going to be removed
  • Assuming files still exist on start-up (see above), alert the user that these files are still there (similar to "New Session or Restart Existing?" dialog)
  • Have the browser pop-up a Save As dialog instead of automatically saving to temp, and use a documents path

Other ideas are possible, and a combination of these might be necessary. Ultimately, this fix should target regular users without much knowledge of the filesystem.

Resources: ted, mfinkle

See also: bug 280419 and bug 369108, and code here.

Standalone Test Harnesses

Mozilla uses several systems for automated testing, including Mochitest, Reftest, xpcshell unit tests, and crash tests. Currently, to run these test suites or develop new tests, you need to build your own Firefox with --enable-tests, and run the tests from the object directory. Ideally there would be a standalone package available for download that would let you run these test suites against any Firefox build, to ease development of new tests and allow more people to get involved with automated testing.

Add source checkout to buildbot

Unlike many open source projects, it's not possible to do a CVS checkout and build the code. This is due to the complex module and version combinations necessary to get and build a particular product, for example Firefox. To accomplish this, Mozilla uses client.mk, which provides a set of rules for getting and building the code.

Increasingly, Mozilla is moving toward using the Python based buildbot system for automated builds and tests. It would be good to add a Source class for client.mk checkouts. This would mean creating a a general Source class that understands the idea of checking out a file, and which could then be used to drive the checkout and allow for integration with client.mk.

References: Ben Hearsum (bhearsum), Rob Helmer (rhelmer)

Thunderbird Image Auto-Resize

Write an extension for Thunderbird that gives functionality similar to that of Outlook, where image attachments in an email can be automatically re-sized to one of a set of smaller sizes. This is helpful for users who would otherwise try to send megabytes of image data, when they can safely scale the images down and still share their pictures with friends.

References: #maildev

Port the Firefox Release Repackager to other platforms

The Firefox Release Repackager allows one to take a released version of the browser, deconstruct it in order to add some custom extension, then repackage it. This allows organizations to ship slightly modified versions of stock releases, useful for their users. Currently the repackager only works on Mac.

Port this tool to Windows and/or Linux. In order to do this you'll need platform binaries of 7zip, or perhaps you can use P7zip to do a cross platform version. There are other ideas that can be added to this tool as well for making it more flexible.

Resources: mkaply, bsmedberg

gdb symbol server support

Microsoft's debugger allows the use of a symbol server, which it can use to download debugging symbols for programs and libraries. Mozilla provides this for nightly and release builds on Windows, and it's useful for people to debug their crashes locally without having to build a debug build. This functionality doesn't exist for Linux or OS X, where gdb is the debugger. gdb does support loading debug symbols from external files, so it could probably be extended to look for the symbols on a web server first.

References:

  1. https://bugzilla.mozilla.org/show_bug.cgi?id=412722#c5
  2. http://www.linuxcommand.org/man_pages/objcopy1.html (see --add-gnu-debuglink and --only-keep-debug)
  3. http://www.totalviewtech.com/Documentation/Tips/gnu_debuglink.html

Resources: ted, Stan Shebs, Jim Blandy

gdb support for minidumps for post-mortem debugging

Microsoft's debugger allows for loading a minidump produced during a crash and then performing post-mortem debugging. GDB supports using core files for the same purpose. In Firefox 3, we produce minidump files on all of our supported platforms for use with our Breakpad crash reporting system. Sometimes people may want to debug a crash locally using one of these minidumps. On Windows, using the symbol server, they can simply load the minidump in their debugger. Currently there is no equivalent on Linux or Mac. Extending gdb to support loading minidumps for post-mortem debugging would be useful. This project may rely on the gdb symbol server project to be truly useful.

References:

  1. http://code.google.com/p/google-breakpad/source/browse/trunk/src/google_breakpad/common/minidump_format.h
  2. http://www.codeproject.com/KB/debug/postmortemdebug_standalone1.aspx
  3. http://sourceware.org/gdb/current/onlinedocs/gdb_3.html#SEC7
  4. http://sourceware.org/gdb/current/onlinedocs/gdb_9.html#SEC80

Resources: ted, Stan Shebs, Jim Blandy

RegressionTry

Make it possible to look for bugs across old nightly builds of Firefox using automated tools. For example, regression-search is a script that allows you to do binary searches across binaries, looking for the existence of a bug. This could be improved through the use of tools like co-scripter. Ideally the user could look for bugs in automated ways, without having to download and install tons of browser versions.

Resources: db48x, ted

Mercurial history browsing

We're just starting to use an exciting new distributed version control tool: Mercurial.

One of the exciting things about Mercurial is that the history—the list of all the changes that have ever been checked in—is not linear. It frequently has branches and merges. This is actually a good, important feature, as you know if you've read a little about distributed version control. The downside is that the history becomes a maze of twisty little passages. Benjamin Smedberg's demo shows the history of a repository with lots of merges (each box is a check-in; you can click on the boxes to move around).

There have been a couple attempts at showing history in an intuitive, graphical way. Mercurial comes with a web UI for browsing the repository, including history (here's what it looks like). It also comes with an "hg glog" extension that draws history as ASCII art, and an "hg view" extension that does roughly the same thing with a little GUI. Sadly, none of these applications qualifies as awesome.

Your mission is to rectify this sad situation by writing an awesome browser-based UI for navigating Mercurial repository history.

Resources: jorendorff, bsmedberg

Extend Try Server to test Thunderbird patches

"Specifically, Mozilla has a great system called “try servers” where one can submit patches against the tree, and the build system runs builds on Linux, Mac and Windows, using those patches, then serves those builds for testing. This is really helpful to figure out if proposed patches solve specific problems...The only problem is that there’s a little bit of patching needed to the try server code itself to make it able to work with other targets besides Firefox, as described in bug 431375" (Ascher)

Add DTrace probes and scripts for Mozilla code base

DTrace was created by Sun in OpenSolaris to allow developers to write simple scripts in order to probe and instrument executing programs in a way not possible otherwise. One of the advantages of DTrace is that it allows you to turn on probing when you need it, but not incur an execution penalty with regard to performance (i.e., when probes aren't being scripted, they aren't run). This requires developers to add "probes" to their source code. For example, if you wanted to know when a particular function is entered/exited, get info about what happened, etc. you could add probes to the particular function. Writing the probes is not difficult, however there are hundreds or thousands of probes that Mozilla would like across its code base.

Resources: sayrer

Plugin Code Test Suite

Create a plugin test suite. There are currently no tests of any kind, and this code is very brittle and not well understood. This would involve creating a plugin or set of plugins and some html/js to drive that plugin in order to push test coverage. Maybe look at the JRE tests, create some specific plugin cases (e.g., swf to load that tests flash specific code). The newly announced Google Chromium browser seems to have something like this with regard to using the npapi for driving tests via plug-in. See also docs on IPC and plug-in design in Chromium.

This will require C programming, as well as light web development (html, js, perhaps swf).

Resources: jst

Add Source Server Support for Mercurial

Previously we added source indexing to release builds using CVS as a backend for pulling source files. Since then, Mozilla has moved to Mercurial from CVS. This project will add support so that Mercurial can be used too. See bug# 440001.

Resources: ted, lsblakk


Create self-serve symbol upload system

Mozilla has a symbol server which is used by developers to debug issues in release builds that don't include debug information (i.e., symbols). However, many crashes are related to third-party binary plugins or extensions, for which Mozilla has no symbols. This project would create a web-based self-serve solution for third-parties to upload symbols for their binary components in breakpad-formatted .sym file format. Ideally this could happen in an automated way, such that third-parties could add a step to their build system to do it automatically with each release. See bug# 419879.

Resources: ted

Add OpenLayers test suite to Mozilla

Mozilla uses many common JavaScript library test suites to add additional coverage to it's JavaScript implementation. One test suite that is not currently used by Mozilla, but could be, is the OpenLayer's js test suite. See bug# 399311.

Resources: sayrer

Import sqlite test suite

Mozilla uses an embedded version of SQLite extensively to store data such as bookmarks, history, etc. To insure that it runs properly within Mozilla, it would be good to add the entire sqlite test suite to Mozilla's tests, so that testing the browser also means testing sqlite. The sqlite test suite is written in TCL and needs to be ported to JavaScript. Some work has already been done, but more is required to get full test coverage.

References: bug 391756, http://www.0xdeadbeef.com/weblog/?p=368

Resources: sdwilsh

Finish Weave/Thunderbird integration

Weave is Mozilla's services project. It currently allows different Firefox installations to synchronize bookmarks, etc. We'd like Thunderbird to be a full-fledged Weave client as well. Some [??? work has already been done], but more is required to get utility. Possible scenarios include:

 * thunderbird accounts synchronized (configure once, use everywhere)
 * automatic bookmarking in Firefox of URLs sent to/from friends
 * access to Thunderbird address book from Firefox 

References: bug 446444

Resources: dmose

Make chromebug work in Thunderbird

TBD

Help gristmill move forward in Thunderbird

TBD

Help Calendar integration on Trunk

TBD

Resources: Fallen

Contribute to Private Browsing tests

The new Private Browsing feature (see bug 248970 and the test plan) needs thorough tests written in order to insure its proper functionality. This will involve collaborating with those writing the patch and tests in order to develop a full suite of tests.

Resources: ehsan, discussion of Private Browsing

Create Content Security Policy test suite

Adding Content Restrictions functionality to Firefox means that sites could restrict types of locations for loaded scripts and other unsafe content. The Content Security Policy is intended to mitigate a large class of Web Application Vulnerabilities: Cross Site Scripting. In order for this to be written properly, a full test suite is required, which can test such things as inline scripts, javascript URLs, all event handler attributes (onclick, etc), plugins, frames, data urls, and XBL.

References: http://people.mozilla.org/~bsterne/content-security-policy/, bug 411791, bug 390910

Resources: bsterne, dveditz, shaver

HTML page set sanitizer

The Talos performance testing system at Mozilla currently runs on a large set of web pages pulled from the Alexa Top 500. These pages can't be redistributed, since they're mirrors of copyrighted web pages. In addition, many of them contain adult content. This makes it difficult for people to duplicate the Talos results or to test changes that have an expected performance impact.

A useful solution to this problem would be a tool that takes a mirrored copy of a website and "sanitizes" it, by changing the page text and image contents (making them junk or filler text or something). The caveat here is that this *cannot* change the performance characteristics of the page. For example, taking a page that is all Chinese text and replacing it with "Lorem Ipsum" filler text would cause the page to take different text rendering paths, which would change what is measured. As another example, making all JPEG images solid black would likely make them decode and render much faster. Any solution should have some analysis performed that shows that performance is not significantly altered in the sanitized page set.

Resources: ted (but find someone better!)


Potential OOo Projects

Sample Project

This is a sample project stub. You can use the template for Sample Project in order to create a project page for one of the stubs below. This is how you 'sign-up' for a project.

NOTE: if someone has already created the project page, speak to this person and see if you can join them. If so, simply add your name to the Project Leader(s) page. Otherwise, you can become a contributor later.

Generate an OOo XREF based on Dehydra

Using Dehydra, build OOo and generate an xref in the same way we are indexing for dxr.

TODO: more info....

Resources: humph, taras