Community:SummerOfCode19

Now that Remote Settings has become a solid solution to ship data, we could use it to publish updates of the PSL. The task consist in migrating the current client code to read from Remote Settings instead of a file, and implement a scheduled job (like Python) to push updates automatically (most likely compiled as a DAFSA file)

i) building additional classifiers (e.g., to detect bugs with no steps to reproduce, or to suggest a developer to assign to a bug, and so on);

ii) improving accuracy/precision/recall of the existing classifiers by implementing other machine learning techniques (e.g., by using convolutional neural networks or recurrent neural networks);

iii) improving accuracy/precision/recall by implementing additional feature extraction steps or making the already existing ones better.

• Detailed Project Description

• A quick test of the memory owned by the crashed process (such as those in memtest86+) might catch faulty memory, a common cause for crashes
• The crash reporter client might try and validate Firefox files which can sometimes become corrupted (usually because of hardware issues). If they are it can prompt the user to re-install Firefox
• Many crashes are caused by buggy graphics drivers, while we blacklist the most egregious offenders we can't cover them all. The crash reporter might identify a faulty driver (if a bug for it has already been filed) and point the user to up-to-date drivers
• For crashes caused by known bugs it should be possible to point the user to the bug filed on bugzilla

-

The ultimate goal of this would be to embed this downloader in our browser for Emerging Markets, Firefox Lite. Firefox Lite is not satisfied with the current Android Download Manager in several ways: We'd like to give users the ability to pause/resume a download, we'd like to download a file directly to SDcard (opposed to download it to the main storage and move it into SDcard). That being said, please focus more on the reusablility of your App as a library than the UX/UI of the App itself. Afterall, GSOC is a coding project!

We've also prepared an even more ambitious mission for those who want tough challenges: Design the app and make it dynamic deliverable. With aabs we can satisfy both light and heavy users by defaulting Android Download Manager as the download tool and prepare the aforementioned downloader dynamically so that Firefox Lite itself is still minimized in terms of disk size. Nevin and mTwTm are Android developers who can provide assistance to Android App and library design.

I would also encourage candidates to differentiate onself by focusing on demonstrating your understanding to Android Apps and download protocols. Brainstorming for more use cases is encouraged but we would really like to learn more about how you can implement this project efficently instead of what you can plan in your proposal.

Also, we're utilizing the Mozilla Asia Product mail-list (https://groups.google.com/forum/#!forum/mozilla-asia-products) for discussions around this project. You might find some useful discussions there as well.

support the occasional website that uses them. Each such format requires the Firefox decoder to use a new open source library for parsing and decoding. This, unfortunately, increases the attack surface of Firefox and as we saw in Pwn2Own 2018, Firefox was successfully exploited via a bugs in such libraries (libogg in this case).

This project proposes to sandbox third-party libraries in Firefox by building a new software-fault isolation toolkit. Our tookit will build on the WebAssembly compiler to isolate libraries in Firefox. But, as part of this toolkit we will also develop and apply a library for safely interfacing with sandboxed libraries (and sanitizing data coming from them). with this toolkit we can ensure that any vulnerability in third-party libraries (e.g., libogg or libpng) cannot be used to be used to compromise Firefox.

The goal of this project is to make sure that tests are executed. We currently have a similar test automation for static analyzer jobs. The idea would be to extend (or replicate) this model for ./mach lint (which run flake8, eslint, etc) and ./mach clang-format The bug is https://bugzilla.mozilla.org/show_bug.cgi?id=1448008

This project would entail updating our security event API to ensure we track and expose the required data. The first phase is to build a script that consumes the API and pretty prints the timeline. The second phase is to provide a web interface for the security timeline.

Writing docs for Firefox's in-tree source docs is time consuming and difficult and the end result is difficult to navigate.

With MDN de-prioritizing build and workflow docs, we need a suitable replacement for all of Firefox's contribution and workflow documentation. The great advantage of documentation living in-tree, is that it can be updated along with the source. Unfortunately the current system to build and generate docs is difficult to write for, slow to build and generates poorly organized documentation. These factors discourage developers from creating or updating docs.

This project aims to improve the documentation experience via static analysis tools (e.g hint when docs might need to be updated), enabling linters, faster build times, additional language support and well structured hierarchies. Help make documentation a bigger part of our developer's day to day workflow.

• C/C++
• Rust
• SIMD

• Scene detection
• Frame type assignment and ordering
• Encoder complexity analysis and design of speed levels
• Rate-distortion optimization algorithms
• Parallelized encoding using Rust fearless concurrency
• Chunked encoding