WineHQ

Summer Of Code: Difference between revisions

No edit summary
(15 intermediate revisions by 4 users not shown)
Line 40: Line 40:
In addition to Google Summer of Code Wine also participates in:
In addition to Google Summer of Code Wine also participates in:
* [https://www.outreachy.org/ Outreachy] is a program similar to GSoC organized by the Software Freedom Conservancy. The goal of Outreachy is to provide encouragement, experience, and networking opportunities for minorities that are underrepresented in tech. Unlike GSoC, it is not limited to students; you can read the Wine Wiki's [[Outreachy]] page for more information.
* [https://www.outreachy.org/ Outreachy] is a program similar to GSoC organized by the Software Freedom Conservancy. The goal of Outreachy is to provide encouragement, experience, and networking opportunities for minorities that are underrepresented in tech. Unlike GSoC, it is not limited to students; you can read the Wine Wiki's [[Outreachy]] page for more information.
* [https://codein.withgoogle.com/ Google Code-In] is aimed at pre-university students and doesn't require any previous coding experience. As a result, the tasks for Code-In participants are more focused, smaller, and have a gentler learning curve. Some aren't even programming tasks but related to other aspects of an open-source project, such as market research, user-support, or documentation. For more details, see the Wine Wiki's [[Code-In]] page.
-----
-----


Line 50: Line 49:
* As long as you work hard and interact with the community and your mentor in a positive and constructive way you don't have to worry about not meeting all your goals.
* As long as you work hard and interact with the community and your mentor in a positive and constructive way you don't have to worry about not meeting all your goals.
----
----
=== Direct3D 10 Effects ===
Possible Mentors: We'll provide you with the appropriate mentor


Difficulty: Medium
=== Fix Tests on Your Windows Machine ===
Possible mentors: Depends on the libraries you pick


* Direct3D 10 implements effects in d3d10.dll. Native d3d10.dll cannot be used in Wine; thus, it would be quite valuable to have a better support for Direct3D 10 effects.
Wine has an extensive testsuite that is used to document how the Windows API works and check Wine's conformance to the native Windows behaviour. The Wine testbot automatically runs tests when patches are submitted to check if a patch breaks anything.
* The work would consists of writing more tests for Direct3D 10 effects and implementing missing functionality.
* The project goal could be set to fix/improve Steam Big Picture rendering in Wine, see bug [https://bugs.winehq.org/show_bug.cgi?id=44120 44120].
----
=== VKD3D - Write more tests for Direct3D 12 ===
Possible Mentors: We'll provide you with the appropriate mentor
 
Knowledge prerequisite: C, knowledge of Direct3D 11, 12 or any 3D graphics API is helpful
 
Difficulty: Easy
 
* [https://source.winehq.org/git/vkd3d.git/ VKD3D] is a D3D12 to Vulkan translation library. It has a test suite which verifies behavior of Direct3D 12 API.
* The work would consists of writing more tests for Direct3D 12 API.
* Direct3D 11 tests from Wine conformance tests are good candidates to be ported to Direct3D 12 tests.
----
=== Direct3D - Microbenchmarks ===
Possible Mentors: [[User:StefanDoesinger|Stefan Dösinger]]
 
Knowledge prerequisite: C, Direct3D, OpenGL
 
Difficulty: Medium


Debugging performance problems in full games is a challenge. Small test programs that test the performance of single operations can help to locate performance problems and fix them. Once fixed, those test programs can be used as regression tests to detect regressions that are hidden inside the margin of error in games. With an equivalent test case in OpenGL it is possible to separate problems in wined3d from problems in the driver.
Unfortunately no Windows machine passes all the tests: http://test.winehq.org/data/ . A few tests are failing reliably and others fail randomly. This can have a number of reasons. Either the test is too strict, Windows' behaviour changed from version to version, the test does not take the influence on some settings into account (e.g. system language), etc.


Stefan Dösinger has already written test programs for draws, buffer uploads, clears and a few other operations and is testing them on a daily basis. More tests need to be written. The test programs will probably be maintained outside the Wine source tree.
A possible GSoC project is to pick a set of libraries of a certain domain you are familiar with (e.g. 3D graphics, XML parsing, networking, etc), where tests are failing on one or more of your machines and try to fix them. However, we don't simply want to remove failing tests, but try to understand why they are not behaving as expected. So be prepared for long debug sessions to find out the differences between your Windows installation and one that passes the tests.


A good application should contain a list of d3d operations to test, a rationale why those operations are good candidates and a short description how each of the tests will be implemented.
Some of the details we expect you to provide in your proposal are DLLs you plan to look at and the current test failures you see in them. Hack away any crashes that prevent any later tests from running to get the full picture. Test how the same tests behave on Linux inside Wine.
----
----
=== Make explorer.exe more useful ===
Possible Mentors: We'll provide you with the appropriate mentor


Knowledge prerequisite: C, Win32
=== Upstream changes from winevdm ===
Possible mentors: Stefan Dösinger


Difficulty: Medium
Wine supports running old Windows 3.x (Win16) programs on 64 bit Linux hosts. The winevdm project (https://github.com/otya128/winevdm) ported this ability to Windows. They are using Wine's 16 bit thunks and combine it with a CPU emulator to get around kernel limitations that prevent 16 bit applications from working natively on 64 bit Windows.


Wine's file explorer allows a user to browse files but many features are missing.  In particular, there is no support for drag & drop copying and moving of files and directories.
The winevdm project has made quite a few changes to the thunks over time and extended them. This project proposal is about upstreaming those changes to Wine. Find some applications that work in winevdm but not Wine, isolate the changes necessary to make them work (the winevdm git log is your friend) and submit them in small patches to wine-devel.


On many platforms, using a native file browser is a more natural alternative. However, on Android (and Chrome OS) machines, improving Wine's file browser could be better than using a native file browser.  This is because of the Android security model and file system structure.  Wine, when installed on Android, lives in a sandbox running as its own user, and on a non-rooted device, a native file browser will not have access to the files inside the Wine prefix.  A user who wants to copy files from the limited areas of shared storage (i.e., the Downloads folder) into the c: drive, or vice versa, basically has to use a wine command prompt.  Adding move / copy features to Wine's file explorer would provide a more user-friendly mechanism (which could also be used on other Wine platforms if desired). 
Your GSoC application should identify one or more Win16 applications together with a likely set of changes that will get the application running. It doesn't have to be a complete set of patches - that will be the main part of your project - but we want to see that you are able to find your way around the source tree.
 
Not being able to interact easily with the wineprefix is a major and basic usability issue with Wine on Android and Chrome OS. You can fix it!
----
----
=== Direct3DRM - Implement rendering for D3DRM ===
Possible Mentors: [[User:StefanDoesinger|Stefan Dösinger]], [[User:AndreHentschel|André Hentschel]]


Knowledge prerequisite: C, DirectDraw, Direct3D
=== Improve HiDPI support ===
 
Difficulty: Hard
 
* D3DRM (Direct3D Retained Mode) is partially implemented, it needs to interact with ddraw to actually display something.
* Even if it's an old library and has been removed in Vista, there are small Games based on it.
* There may be value in making this work on Windows / native ddraw as well.
* Some bugs with games that use d3drm: [https://bugs.winehq.org/show_bug.cgi?id=12851 12851], [https://bugs.winehq.org/show_bug.cgi?id=19733 19733], [https://bugs.winehq.org/show_bug.cgi?id=21670 21670]. There are more of them.
* Basic object creation, ddraw initialization and stand-alone math functions are implemented. Very rudimentary display functionality exists.
* The most notable missing features are geometry and texture handling, state management and actual 3D draws.
Applicants should have understanding of how 3D APIs work, and specific knowledge in old Direct3D / DirectDraw versions will be of an advantage. You will have to work with file formats for 3D geometry and textures, fixed function transformation and the really awful Direct3D 1 interface.
----
 
=== D3DX9 - Implement missing D3DX9 APIs ===
Possible Mentors: [[User:StefanDösinger|Stefan Dösinger]], [[User:MatteoBruni|Matteo Bruni]]
 
Knowledge prerequisite: C, maybe some domain-specific knowledge related to the APIs you want to implement
 
Difficulty: Easy to medium, depending on the functions you want to work on
 
* [[DirectX-D3DX9|D3DX9 dlls]] contain various utility functions related to common 3D graphics tasks. For instance they contain math functions, functions for loading textures, compiling shaders and much more.
* A good chunk of those APIs has been implemented but a lot is still missing, such as:
** some math functions (what's left is mostly about spherical harmonics and precomputed radiance transfer)
** some mesh functions (D3DXComputeNormals(), D3DXOptimizeVertices(), ...)
** a few texture functions
** quite a few ID3DXConstantTable methods
** effect framework bits and pieces
** drawing text in ID3DXFont
* The project would choose a reasonable subset of the missing functions and the work would consist of writing behavioral tests for those functions and actually implementing the functions themselves. We will help you in selecting a reasonably-sized subset of the API, suitable for a GSoC project.
----
=== Tools - Winetest Scripting Interface ===
Possible mentors: [[User:StefanDoesinger|Stefan Dösinger]]
 
Knowledge prerequisite: C, scripting language of your choice
 
Difficulty: Easy
 
Wine has an extensive set of regression tests. These tests can be helpful for projects Wine depends on. E.g. the developers of 3D drivers may want to run our d3d tests and our d3d implementation to test their OpenGL implementations.
 
One obstacle Mesa developers reported is that the default way to run tests (run "make test") only reports success or failure. That's OK when all tests are passing, but it is a problem when there are known test failures and you want to make sure you don't introduce additional failures. In this case users have to manually check the output written to stdout, which is a terrible task.
 
The Wine testbot has some scripts to ignore known failures. It might help developers of dependency libraries if we made them available outside the testbot environment somehow. On the other hand we do not want to make the existing test code more complicated than it is. Ideally the tests themselves would not be modified.
 
For your proposal please think about what the API of such a scripting interface would look like and which information it provides to its callers.
----
=== DirectWrite: Implement vertical text layout support ===
Possible mentors: [[User:NikolaySivov|Nikolay Sivov]]
 
Knowledge prerequisite: C
 
Difficulty: Medium
 
DirectWrite is a text layout and rendering library available on recent Windows releases, starting with Vista. It provides several API layers to manage font file, process text, render it, and interact with GDI subsystem. A lot of basic functionality is already supported in Wine, this proposal is however about specific feature to process and render text in vertical mode. This touches every layer mentioned above, brief summary on what needs to be done follows:
 
* AnalyzeVerticalGlyphOrientation - this method analyzes input text and returns orientation for every part of the text;
* GetVerticalGlyphVariants - returns vertical glyphs for specified nominal glyphs, if font has any;
* vertical glyph metrics should be properly returned, that includes GetDesignGlyphAdvances;
* layout part has to support vertical direction when building lines, this implies baseline calculation, alignment logic changes, maybe something else;
* IDWriteBitmapRenderTarget1 and IDWriteGlyphRunAnalysis have to support sideways mode too.
----
=== crypt32: Implement DSS (digital signature standard) Provider ===
Possible mentors: Hans Leidekker
 
Knowledge prerequisite: C
 
Difficulty: Medium
 
Many .NET applications fail to run, because currently the DSS and enhanced DSS provider are missing in crypt32.dll.
There are already some tests for that, but more might be necessary. See also [https://bugs.winehq.org/show_bug.cgi?id=10506 Bug 10506]
----
 
=== AppDB improvements ===
Possible mentors: Jeremy Newman, Rosanne DiMesio
 
Knowledge prerequisite: PHP, HTML, MySQL
 
Wine's Application Database (https://appdb.winehq.org/) is old and could use improvements. If you have ideas for improvements and new features please step forward!
----
=== Uniscribe: Improve Indic language support ===
Possible mentors: [[User:AricStewart|Aric Stewart]]
 
Knowledge prerequisite: C, familiarity with fonts and opentype a plus,  familiarity/fluency in Devanagari, Bengali, Gurmukhi, Gujarati, Oriya, Tamil, Telugu, Kannada or Malayalam not required but a big plus.
 
Difficulty: Medium
 
Uniscribe is the text layout and rendering system for earlier versions of windows and much of the Wine UI.  The frameworks is mostly in place as well as a first pass implementation for many languages including the Indic languages. However nobody with more than a passing knowledge of the languages has dug into the details and issues in properly displaying these complicated languages.  Including issues around separate Indic matras. Ideally the student working on this will also be fluent enough to at least start doing some translation of the Wine strings into the target language as well, so that Wine's usability in those languages greatly improves.
----
=== Implement Windows Media Player control ===
Possible mentors: We'll provide you with the appropriate mentor
Possible mentors: We'll provide you with the appropriate mentor


Knowledge prerequisite: C, COM+, OLE32, DirectShow?
Wine currently supports HiDPI. But there're many places in Wine UI that don't take HiDPI into consideration. An example application is Control Spy v6 2.0 with DPI set to 192. DateTime Picker doesn't have enough space, same for Edit, Hot Key, Month Calendar, Status Bar, SysLink, Tab Control, Trackbar, and Up/Down. Note that it may be an application bug. But there are other places in Wine apps as well, e.g., print preview button and toolbar size in WordPad. First, get a scope of Wine common controls/user controls/applications that need to be fixed. Then for further improvements, you may work on setting the recommended DPI scale at Wine startup or dynamic DPI change at runtime.
 
Difficulty: Medium
 
A lot of apps and libraries use Windows Media Player COM/ActiveX control to implement media playback (Including MS .Net's  System.Windows.Media.MediaPlayer class).
Users can usually workaround this by just installing windows media player 9 or 10. However this technically requires a Windows license.
Wine currently has a stub-only implementation of WMP.
WMP can be used in a headless mode for media playback or could be used as a full UI component with full controls. For the purpose of GSoC scope could be limited to headless parts.
 
It would be nice to have a builtin implementation.  Potentially media playback could be implemented using DirectShow.
 
This project should most likely have an end goal of  "Have <insert app/game here> work with builtin WMP implementation and have audio/video media playback working".
----
----
=== Winecfg / winemenubuilder - enhance MIME type handling ===
Possible Mentors: [[User:MichaelMueller|Michael Müller]], [[User:SebastianLackner|Sebastian Lackner]]
Knowledge prerequisite: C
Difficulty: Medium


Wine tries to integrate Windows programs as much as possible into your unix/linux system and provides ways to directly assign MIME types to programs running in Wine. This makes it possible to open Word when you click on a .docx file in your file browser. However, this kind of integration is not always desirable, especially if there are native programs which support the same MIME type. At the moment a user can only decide to disable this kind of integration completely or Wine will automatically forward all MIME type registrations of Windows programs, resulting in questionable ones like .txt -> notepad and .png -> Wine Internet Explorer. If a user wants to remove such MIME type registration again, it is necessary to manually delete the according files.
=== Multi-monitor display settings support on Mac ===
Possible mentors: Zhiyi Zhang


The idea of this task is to provide a GUI (most probably as part of winecfg) to control the creation of such MIME type assignments and therefore making Wine more user friendly. You can find ideas how such a GUI could look like in the [https://forum.winehq.org/viewtopic.php?f=2&t=24237 Forum] or in the [https://bugs.winehq.org/show_bug.cgi?id=19182 Bug Tracker]. This task is also suitable for new Wine developers as no deep knowledge about the Wine source code is required. It might help though to have some knowledge about Win32 Dialogs.
* Currently, Wine supports enumerating multiple monitors(EnumDisplayMonitors) but cannot change their display settings on Mac(EnumDisplaySettings, ChangeDisplaySettings). On Mac, only the primary display settings can be changed. Multi-monitor display setting handlers are implemented in winex11.drv so we can do a similar thing in winemac.drv.  
* End goal would be to allow changing the display setting on multi-monitor Mac systems and passing all the tests in dlls/user32/tests/monitor.c
----
----
=== XMLLite: Implement fully compatible xmllite parser ===
Possible mentors: [[User:NikolaySivov|Nikolay Sivov]]


Knowledge prerequisite: C
=== Implement XRandR display settings handler with transforms ===
Possible mentors: Zhiyi Zhang


Difficulty: Medium
* On Wayland and some setups with Nvidia proprietary drivers, XRandR reports only one resolution(https://bugs.winehq.org/show_bug.cgi?id=34348#c34). In this case, the proper solution is to use XRandR transforms to simulate display resolution changes so that Wine can support multiple display resolutions even if the host only report the native resolution.
 
* End goal would be to allow changing the display setting on Wayland and passing all the tests in dlls/user32/tests/monitor.c
* There shoudn't be any external dependencies for parsing functionality
* The main focus should be on the Reader
* Full parser resuming functionality for pending reads
* Unit tests
* Support for external entities to allow nested parser inputs
----
----
=== Tools - Implement resource editor and / or dialog editor ===
Possible Mentors: [[User:MarcusMeissner|Marcus Meissner]]
=== DirectShow DVD support ===
Possible mentors: Zebediah Figura


Knowledge prerequisite: C, some UI programming perhaps
* DirectShow is a general multimedia streaming framework, built around creating graphs of individual "filters" which consume and produce data. Windows ships a number of built-in filters.
 
* Several interfaces related to DVD support (DVD Navigator filter, DVD filter graph) are stubs or unimplemented. The work would consist of providing an implementation of these interfaces.
Difficulty medium, might be bit large for a student project
* End goal would be to allow a native application to play DVDs.
 
* We currently have no graphical dialog / resource editor
* Free external ones are AdWare infected or Payware
----
----


=== CMD - implement more robust parser ===
=== AVFoundation video capture support ===
Possible Mentors: We'll provide you with the appropriate mentor
Possible mentors: Zebediah Figura, Gijs Vermeulen<br>
 
Requirements: Mac OS development
Knowledge prerequisite: C, bison, flex, CMD commands
 
Difficulty: Medium


* http://kegel.com/wine/sweng/2010/ is a project to fix a number of small bugs in Wine's cmd.exe and friends, and to write testcases for it. It has helped bring to light certain flaws in cmd.exe's parsing of if/then/else and () blocks.
* Yet another DirectShow project: implement video capture on Mac.
* Somebody could quite usefully spend a summer bulking up the test cases for cmd and friends, fixing any remaining reported bugs, and improving the parser.
* Currently the only video capture backend is Video4Linux (i.e. v4l2). The work would consist of writing another backend using AVFoundation.
* End goal would be to allow a native application to capture video.
----
----


=== Portability - Port WineLib to a new architecture ===
=== DirectShow audio capture ===
Possible mentors: [[User:AndreHentschel|André Hentschel]]
Possible mentors: Zebediah Figura
 
Knowledge prerequisite: Assembler for the target
 
Difficulty: Medium, depending on the target


* Wine can run WineLib applications on different CPU architectures, but needs at least some assembler code for that.
* Yet another DirectShow project: implement audio capture.
* Suggested architectures: [https://en.wikipedia.org/wiki/Ppc64 PowerPC64 (maybe POWER8?)], [http://en.wikipedia.org/wiki/Sparc64 Sparc64], [https://en.wikipedia.org/wiki/RISC-V RISC-V].
* The work would consist of implementing the CLSID_AudioRecord object, which is currently only stubs.
* You will also need a test system. Qemu should be possible.
* Audio capture should probably be done through WinMM APIs (i.e. waveIn*).
* End goal would be to allow a native application to capture audio.
----
----


<!--
=== Implement robocopy.exe ===
=== VFW - Implement Indeo codecs ===
Possible mentors: Zebediah Figura
Possible mentors: [[User:zf|Zebediah Figura]]


Knowledge prerequisite: C
* A complex copying program needed by some installers: <https://bugs.winehq.org/show_bug.cgi?id=43653>
 
* Work would consist of implementing the basic functionality of the program, including flags needed by the relevant programs.
Difficulty: Easy
 
* Many old games need an implementation of the various Indeo codecs (ir32_32.dll, etc.) For an example see [https://bugs.winehq.org/show_bug.cgi?id=27244 bug 27244].
* Actual decompression code should probably be copied from libavcodec.
----
----
-->
<!--
=== Implement calculator app ===
Possible mentors:


Knowledge prerequisite: C
=== Evaluate performance of hqemu in Hangover ===
Possible mentors: Stefan Dösinger


Difficulty: Easy
Hangover (https://github.com/AndreRH/hangover) is a proof of concept of integrating a CPU emulator - in this case qemu - with Wine to run x86 Windows applications on non-x86 host CPUs without running a full Linux userspace stack inside the emulator. While its design is not suitable for upstream integration, it is useful for performance evaluation.


* There are actually multiple applications which try to open the Windows Calculator, see [https://bugs.winehq.org/show_bug.cgi?id=15555 bug 15555].
hqemu (http://csl.iis.sinica.edu.tw/hqemu/) is a modification of qemu to generate more efficient translated code with the help of LLVM.
* [http://www.jebriggs.com/winecalc/ GNU winecalc] is an open source (LGPL) calculator app for Win32, but it would require so much refactoring to be accepted into Wine that it would probably be easier to start from scratch.
* ReactOS also has an open source calculator app, but its stricter GPL license means that it cannot be used in Wine in any form.
* The first step would be to implement a 4-function calculator, and then add features like memory registers and a scientific mode.
----
-->
 
=== mspatcha - implement patch parser ===
Possible mentors: [[User:zf|Zebediah Figura]]
 
Knowledge prerequisite: C
 
Difficulty: Medium
 
* Many MSI installers ship patch files to update or install programs; see [https://bugs.winehq.org/show_bug.cgi?id=12501 bug 12501]. These patch files are in an undocumented binary format.
* Project would consist of reverse-engineering the patch format and implementing a parser. Ultimate goal is a working ApplyPatchToFile() implementation.
* A way to create new patch files (i.e. mspatchc.dll / mpatch.exe) could also be useful.
 
----
=== Testbot ideas ===
See the [https://testbot.winehq.org/ Wine Test Bot] site and the source at [https://source.winehq.org/git/tools.git/tree/HEAD:/testbot tools.git]
==== Javascript-ification of the site ====
Possible mentors: François Gouget
 
Knowledge prerequisites: Perl (server side), JavaScript, REST or equivalent
 
The goal would be to avoid reloading pages all the time, starting with the Job details page and moving on to other pages if time permits.
* A prime target would be the job details page which reloads every time one shows / hides a log or a screenshot. Dealing with screenshots should be easy. The interesting part would be to have the website provide an API to retrieve the logs which the JavaScript code would then insert in the right place, taking into account that logs are not available until a task has completed. Also, instead of reloading the whole page regularly, the JavaScript code should query for changes to the Job status and update the page accordingly.
* Similarly, instead of reloading all of the main page regularly, JavaScript could be used to retrieve just the new jobs and status changes, and then update the page.
* The job submit page could also use JavaScript to have all operations happen on a single page and make uploading the patch / executable smoother.
* Finally, refreshing the activity page could benefit from a similar treatment to the main page.
* Note: As part of this project you will have to set up your own TestBot instance, that is the Engine, webserver and a QEmu VM.
 
----
==== Swap TestBot's Object Relational Mapper ====
Possible mentors: François Gouget
 
Knowledge prerequisite: Perl, SQL
 
* The TestBot uses its own Perl ORM implementation which has a lot of issues: [https://bugs.winehq.org/show_bug.cgi?id=45023 Bug 45023]
* These issues are currently not blocking the TestBot development but the performance is pretty bad: see for instance the load times for the main and activity pages. It seems likely that the current ORM is responsible for the poor performance.
* So the goal of this project would be to switch the TestBot to a standard ORM, one with better performance and that does not need to be maintained as part of the TestBot codebase.
* The first task will be to pick a replacement Perl ORM based on functionality, community support and performance. The next point can help for that last part.
* To test the performance aspect, the project can convert access to the Records and RecordGroups tables since they are isolated.
* The difficulty with the switch is that the rest of the TestBot's tables are all connected together: in $Task->VM the ORM takes care of returning the appropriate VM object for the Task, which would not work if the Task and VM tables are accessed using different ORMs. So an important part of the project will be to figure out how to handle the transition.
* Note: As part of this project you would have to set up your own TestBot instance, that is the Engine, webserver and a QEmu VM.
 
----
==== Implement load balancing and fail-over ====
Possible mentors: François Gouget
 
Knowledge prerequisite: Perl, SQL
 
Difficulty: Hard
 
See [https://bugs.winehq.org/show_bug.cgi?id=39412 Bug 39412]
This implies:
* Currently a VM, such as a Windows 7 VM, can only be present on a single VM host. If that host goes down, the corresponding tasks cannot be run: there is no fail-over. Similarly if the VM host is busy running other VMs, the task cannot be moved to a less busy host: there is no load balancing. The goal is to change that to improve the reliability and performance of the TestBot.
* In the current schema a Task is associated to the VM that it will run on. However that model will need to be expanded:
** We can already have multiple test configurations for a single VM. This is done by having one VM snapshot per configuration, for instance one per locale. Since there is a single VM, only one configuration can be tested at a time. This is currently represented as separate VM objects and it's up to the task scheduler to detect when two objects reference the same VM and thus to not try to use both at the same time. The new schema should probably find a better way to represent the relationship between the tasks, test configurations and VMs.
** The VM object specifies the hypervisor instance, and thus VM host, used to access it. If a VM is present on multiple VM hosts it's possible to create one VM object for each. However since Tasks link directly to VMs, once created a task would be linked to the VM instance on a specific VM hosts, thus preventing failover. So a better are of representing the relationship between the tasks, VMs, VM instances and VM host / hypervisor is needed.
** The new schema will also need a way to represent licensing restrictions. The TestBot typically has a single license for the Windows VMs, which means at most one instance can run at a time, even if the VM is present on two hosts. But the TestBot should be able to run two or more instances of the Linux-based VMs simultaneously.
** While allocating hypervisor resources for specific VM instances is not part of this project, the new schema should be extensible so one can represent the hypervisor resources and the corresponding VM requirements.
* So the project will need to:
** Update the database schema to account for the above requirements.
** Provide SQL code to update the existing tables.
* Update the code impacted by the schema changes (essentially everything).
* Update the task scheduler to implement fail-over and load balancing.
* Note: As part of this project you would have to set up your own TestBot instance, that is the Engine, webserver and a couple of QEmu VMs.
 
----
=== Unsorted further ideas ===
==== Further cmd.exe improvements ====
 
Further work on the cmd.exe implementation. We need to scope / find out what is missing before offering.
 
----
==== Fix Windows theming / implement themes ====
 
After we fix Windows theming in Wine (project 1) we could implement themes (project 2).
 
----
==== Improve Wine's fonts ====
 
Improve our own fonts and/ or add new fonts.
(This is not for GSOC usually, more for Outreachy)
 
----
==== Improve accessibility of Wine ====
 
Wines accessibility for e.g. blind people is lacking and can use improvements.
 
----
==== winetest improvements ====
 
Improve the server and/or client side of the wine test framework. (?)
 
----
==== signtool.exe ====
 
A tool to verify signatures on MSI packages / EXEs(?)
 
(Need to evaluate if effort matches GSOC effort scope.)
 
----
==== tasklist.exe ====
 
A tool to list currently running programs.
 
(Need to evaluate if effort matches GSOC effort scope.)
 
See, e.g., https://bugs.winehq.org/show_bug.cgi?id=43990
 
----
==== DPI Resolution handling ====
 
We have some DPI resolution issues, some work in this area is needed.
 
----
==== DirectPlay ====
 
Is there any work needed on DirectPlay... Identify missing things.
 
----
==== Common Item Dialog Box improvements ====
 
Ask Vincent Povirk about details / ideas.
 
Improvements of this dialog box framework.
 
----
==== Improvements of specific Windows applications ====
 
Hard to scope the effort for specific applications beforehand to meet the GSOC effort range.
 
(brainstorm: Perhaps select some "known" apps that might be in the range?)
 
----
==== Tools - implement a wine diagnostic tool ====


Collect data about the current installation that can be used for AppDB or bugreporting. Things like wine version, operating system, native dlls versions, etc.
This project proposal is about merging together the hqemu modifications and hangover qemu modifications to evaluate the combined performance. Hangover ideally keeps the amount of code that needs to be emulated small and hqemu speeds up the emulation. If combined, they might either synergize well, or the faster emulation makes hangover's high level thunks moot.


To work on this you will need an arm64 Linux machine. Hangover is notoriously difficult to build. Please try to build and run both Hangover and hqemu separately before submitting your application. Try to get an idea of the modifications each project makes to qemu and how to reconcile them.
----
----


[[Category:Development]] [[Category:Internships]] [[Category:ToDo]]
[[Category:Development]] [[Category:Internships]] [[Category:ToDo]]

Revision as of 08:32, 14 June 2021

Wine and the Google Summer of Code

This page collects some tips for students who want to work on Wine during the Summer of Code, and provides a few ideas for projects.

To apply, go to the Google Summer of Code home page.


Important notes!

  • Applicants MUST have submitted a patch or testcase to wine-devel (see Submitting Patches) to be considered for acceptance.
  • Please first send a draft proposal and discuss it with us, don't send the final one directly.
  • We cannot accept ReactOS proposals, so please only make proposals that will benefit Wine.

To apply, go to the Google Summer of Code home page.


Beware of Legal Requirements

You must state that you will follow these minimal legal requirements during the SoC (and have done so in the past):

  • You are not allowed to read or reuse Windows source code (leaked source / Windows Research Kernel* / ...)

(* we are following the SFLC's advice)

  • You are not allowed to reverse engineer Windows files by disassembling or decompiling them
  • You are not allowed to analyze Windows files with the trace functions of Wine
  • People who work or have worked for Microsoft should probably not participate

Other Outreach Programs

In addition to Google Summer of Code Wine also participates in:

  • Outreachy is a program similar to GSoC organized by the Software Freedom Conservancy. The goal of Outreachy is to provide encouragement, experience, and networking opportunities for minorities that are underrepresented in tech. Unlike GSoC, it is not limited to students; you can read the Wine Wiki's Outreachy page for more information.

Ideas

Your own idea

Possible mentors: We'll provide you with the appropriate mentor

  • If you have an idea, please post it on Wine Developers mailing list so we can help you with your idea and find out if it's realistic or not. Showing initiative and willing to discuss your idea greatly improves your chances of getting accepted. Even more so than taking one of the ideas below.
  • As long as you work hard and interact with the community and your mentor in a positive and constructive way you don't have to worry about not meeting all your goals.

Fix Tests on Your Windows Machine

Possible mentors: Depends on the libraries you pick

Wine has an extensive testsuite that is used to document how the Windows API works and check Wine's conformance to the native Windows behaviour. The Wine testbot automatically runs tests when patches are submitted to check if a patch breaks anything.

Unfortunately no Windows machine passes all the tests: http://test.winehq.org/data/ . A few tests are failing reliably and others fail randomly. This can have a number of reasons. Either the test is too strict, Windows' behaviour changed from version to version, the test does not take the influence on some settings into account (e.g. system language), etc.

A possible GSoC project is to pick a set of libraries of a certain domain you are familiar with (e.g. 3D graphics, XML parsing, networking, etc), where tests are failing on one or more of your machines and try to fix them. However, we don't simply want to remove failing tests, but try to understand why they are not behaving as expected. So be prepared for long debug sessions to find out the differences between your Windows installation and one that passes the tests.

Some of the details we expect you to provide in your proposal are DLLs you plan to look at and the current test failures you see in them. Hack away any crashes that prevent any later tests from running to get the full picture. Test how the same tests behave on Linux inside Wine.


Upstream changes from winevdm

Possible mentors: Stefan Dösinger

Wine supports running old Windows 3.x (Win16) programs on 64 bit Linux hosts. The winevdm project (https://github.com/otya128/winevdm) ported this ability to Windows. They are using Wine's 16 bit thunks and combine it with a CPU emulator to get around kernel limitations that prevent 16 bit applications from working natively on 64 bit Windows.

The winevdm project has made quite a few changes to the thunks over time and extended them. This project proposal is about upstreaming those changes to Wine. Find some applications that work in winevdm but not Wine, isolate the changes necessary to make them work (the winevdm git log is your friend) and submit them in small patches to wine-devel.

Your GSoC application should identify one or more Win16 applications together with a likely set of changes that will get the application running. It doesn't have to be a complete set of patches - that will be the main part of your project - but we want to see that you are able to find your way around the source tree.


Improve HiDPI support

Possible mentors: We'll provide you with the appropriate mentor

Wine currently supports HiDPI. But there're many places in Wine UI that don't take HiDPI into consideration. An example application is Control Spy v6 2.0 with DPI set to 192. DateTime Picker doesn't have enough space, same for Edit, Hot Key, Month Calendar, Status Bar, SysLink, Tab Control, Trackbar, and Up/Down. Note that it may be an application bug. But there are other places in Wine apps as well, e.g., print preview button and toolbar size in WordPad. First, get a scope of Wine common controls/user controls/applications that need to be fixed. Then for further improvements, you may work on setting the recommended DPI scale at Wine startup or dynamic DPI change at runtime.


Multi-monitor display settings support on Mac

Possible mentors: Zhiyi Zhang

  • Currently, Wine supports enumerating multiple monitors(EnumDisplayMonitors) but cannot change their display settings on Mac(EnumDisplaySettings, ChangeDisplaySettings). On Mac, only the primary display settings can be changed. Multi-monitor display setting handlers are implemented in winex11.drv so we can do a similar thing in winemac.drv.
  • End goal would be to allow changing the display setting on multi-monitor Mac systems and passing all the tests in dlls/user32/tests/monitor.c

Implement XRandR display settings handler with transforms

Possible mentors: Zhiyi Zhang

  • On Wayland and some setups with Nvidia proprietary drivers, XRandR reports only one resolution(https://bugs.winehq.org/show_bug.cgi?id=34348#c34). In this case, the proper solution is to use XRandR transforms to simulate display resolution changes so that Wine can support multiple display resolutions even if the host only report the native resolution.
  • End goal would be to allow changing the display setting on Wayland and passing all the tests in dlls/user32/tests/monitor.c

DirectShow DVD support

Possible mentors: Zebediah Figura

  • DirectShow is a general multimedia streaming framework, built around creating graphs of individual "filters" which consume and produce data. Windows ships a number of built-in filters.
  • Several interfaces related to DVD support (DVD Navigator filter, DVD filter graph) are stubs or unimplemented. The work would consist of providing an implementation of these interfaces.
  • End goal would be to allow a native application to play DVDs.

AVFoundation video capture support

Possible mentors: Zebediah Figura, Gijs Vermeulen
Requirements: Mac OS development

  • Yet another DirectShow project: implement video capture on Mac.
  • Currently the only video capture backend is Video4Linux (i.e. v4l2). The work would consist of writing another backend using AVFoundation.
  • End goal would be to allow a native application to capture video.

DirectShow audio capture

Possible mentors: Zebediah Figura

  • Yet another DirectShow project: implement audio capture.
  • The work would consist of implementing the CLSID_AudioRecord object, which is currently only stubs.
  • Audio capture should probably be done through WinMM APIs (i.e. waveIn*).
  • End goal would be to allow a native application to capture audio.

Implement robocopy.exe

Possible mentors: Zebediah Figura


Evaluate performance of hqemu in Hangover

Possible mentors: Stefan Dösinger

Hangover (https://github.com/AndreRH/hangover) is a proof of concept of integrating a CPU emulator - in this case qemu - with Wine to run x86 Windows applications on non-x86 host CPUs without running a full Linux userspace stack inside the emulator. While its design is not suitable for upstream integration, it is useful for performance evaluation.

hqemu (http://csl.iis.sinica.edu.tw/hqemu/) is a modification of qemu to generate more efficient translated code with the help of LLVM.

This project proposal is about merging together the hqemu modifications and hangover qemu modifications to evaluate the combined performance. Hangover ideally keeps the amount of code that needs to be emulated small and hqemu speeds up the emulation. If combined, they might either synergize well, or the faster emulation makes hangover's high level thunks moot.

To work on this you will need an arm64 Linux machine. Hangover is notoriously difficult to build. Please try to build and run both Hangover and hqemu separately before submitting your application. Try to get an idea of the modifications each project makes to qemu and how to reconcile them.


This page was last edited on 14 June 2021, at 08:32.