1# Contributing 2 3Feedback and contributions are very welcome! 4 5Here's help on how to make contributions, divided into the following sections: 6 7The quick-read part: 8* general information, 9* vulnerability reporting, 10* documentation. 11 12The long-read part: 13* code changes, 14* how to check proposed changes before submitting them, 15* reuse of other libraries, frameworks etc. 16 17## General information 18 19For specific proposals, please provide them as 20[pull requests](https://github.com/quotient-im/libQuotient/pulls) 21or 22[issues](https://github.com/quotient-im/libQuotient/issues) 23For general discussion, feel free to use our Matrix room: 24[#quotient:matrix.org](https://matrix.to/#/#quotient:matrix.org). 25 26If you're new to the project (or FLOSS in general), 27[issues tagged as easy](https://github.com/quotient-im/libQuotient/labels/easy) 28are smaller tasks that may typically take 1-3 days. 29You are welcome aboard! 30 31### Pull requests and different branches recommended 32 33Pull requests are preferred, since they are specific. 34See the GitHub Help [articles about pull requests](https://help.github.com/articles/using-pull-requests/) 35to learn how to deal with them. 36 37We recommend creating different branches for different (logical) 38changes, and creating a pull request when you're done into the master branch. 39See the GitHub documentation on 40[creating branches](https://help.github.com/articles/creating-and-deleting-branches-within-your-repository/) 41and 42[using pull requests](https://help.github.com/articles/using-pull-requests/). 43 44### How we handle proposals 45 46We use GitHub to track all changes via its 47[issue tracker](https://github.com/quotient-im/libQuotient/issues) and 48[pull requests](https://github.com/quotient-im/libQuotient/pulls). 49Specific changes are proposed using those mechanisms. 50Issues are assigned to an individual who works on it and then marks it complete. 51If there are questions or objections, the conversation area of that 52issue or pull request is used to resolve it. 53 54<!-- 55### Developer Certificate of Origin (DCO) - not enforced yet 56 57All contributions (including pull requests) must agree to 58the [Developer Certificate of Origin (DCO) version 1.1](doc/dco.txt). 59This is exactly the same one created and used by the Linux kernel developers 60and posted on <http://developercertificate.org/>. 61This is a developer's certification that he or she has the right to 62submit the patch for inclusion into the project. 63 64Simply submitting a contribution implies this agreement, however, 65please include a "Signed-off-by" tag in every patch 66(this tag is a conventional way to confirm that you agree to the DCO). 67You can do this with <tt>git commit --signoff</tt> (the <tt>-s</tt> flag 68is a synonym for <tt>--signoff</tt>). 69 70Another way to do this is to write the following at the end of the commit 71message, on a line by itself separated by a blank line from the body of 72the commit: 73 74 Signed-off-by: YOUR NAME <YOUR.EMAIL@EXAMPLE.COM> 75 76You can signoff by default in this project by creating a file 77(say "git-template") that contains 78some blank lines and the signed-off-by text above; 79then configure git to use that as a commit template. For example: 80 81 git config commit.template ~/cii-best-practices-badge/git-template 82 83It's not practical to fix old contributions in git, so if one is forgotten, 84do not try to fix them. We presume that if someone sometimes used a DCO, 85a commit without a DCO is an accident and the DCO still applies. 86--> 87### License 88 89Unless a contributor explicitly specifies otherwise, we assume contributors 90to agree that all contributed code is released either under *LGPL v2.1 or later*. 91This is more than just [LGPL v2.1 libQuotient now uses](./COPYING) 92because the project plans to switch to LGPL v3 for library code in the near future. 93<!-- The below is invalid yet! 94All new contributed material that is not executable, including all text when not executed, is also released under the 95[Creative Commons Attribution 4.0 International (CC BY 4.0) license](https://creativecommons.org/licenses/by/4.0/) or later. 96--> 97 98Any components proposed for reuse should have a license that permits releasing 99a derivative work under *LGPL v2.1 or later* or LGPL v3. Moreover, the license of 100a proposed component should be approved by OSI, no exceptions. 101 102## Vulnerability reporting (security issues) - see [SECURITY.md](./SECURITY.md) 103 104## Documentation changes 105 106Most of the documentation is in Markdown format. All Markdown files use the .md 107filename extension. Any help on fixing/extending these is more than welcome. 108 109Where reasonable, limit yourself to Markdown that will be accepted by different 110markdown processors (e.g., what is specified by CommonMark or the original 111Markdown). In practice, as long as libQuotient is hosted at GitHub, 112[GFM (GitHub-flavoured Markdown)](https://help.github.com/articles/github-flavored-markdown/) 113is used to show those files in a browser, so it's fine to use its extensions. 114In particular, you can mark code snippets with the programming language used; 115blank lines separate paragraphs, newlines inside a paragraph do *not* force a line break. 116 117Beware - this is *not* the same markdown algorithm used by GitHub when it 118renders issue or pull comments; in those cases 119[newlines in paragraph-like content are considered as real line breaks](https://help.github.com/articles/writing-on-github/); 120unfortunately this other algorithm is *also* called GitHub-flavoured markdown. 121(Yes, it'd be better if there were different names for different things.) 122 123In your markdown, please don't use tab characters and avoid "bare" URLs. 124In a hyperlink, the link text and URL should be on the same line. 125While historically we didn't care about the line length in markdown texts 126(and more often than not put the whole paragraph into one line), this is subject 127to change anytime soon, with 80-character limit _recommendation_ 128(which is softer than the limit for C/C++ code) imposed on everything 129_except hyperlinks_ (because wrapping hyperlinks breaks the rendering). 130 131Do not use trailing two spaces for line breaks, since these cannot be seen 132and may be silently removed by some tools. If, for whatever reason, a blank line 133is not an option, use <tt><br /></tt> (an HTML break). 134 135## End of TL;DR 136 137If you don't plan/have substantial contributions, you can end reading here. Further sections are for those who's going to actively hack on the library code. 138 139## Code changes 140 141The code should strive to be DRY (don't repeat yourself), clear, and obviously 142correct (i.e. buildable). Some technical debt is inevitable, 143just don't bankrupt us with it. Refactoring is welcome. 144 145### Code style and formatting 146 147As of Quotient 0.6, the C++ standard for newly written code is C++17, with a few 148restrictions, notably: 149* standard library's _deduction guides_ cannot be used to lighten up syntax 150 in template instantiation, i.e. you have to still write 151 `std::array<int, 2> { 1, 2 }` instead of `std::array { 1, 2 }` or use helpers 152 like `std::make_pair` - once we move over to the later Apple toolchain, this 153 will be no more necessary. 154* enumerators and slots cannot have `[[attributes]]` because moc of Qt 5.9 155 chokes on them. This will be lifted when we move on to Qt 5.12 for the oldest 156 supported version. 157* things from `std::filesystem` cannot be used yet until we push the oldest 158 required g++/libc to version 8. 159 160The code style is defined by `.clang-format`, and in general, all C++ files 161should follow it. Files with minor deviations from the defined style are still 162accepted in PRs; however, unless explicitly marked with `// clang-format off` 163and `// clang-format on`, these deviations will be rectified any commit soon 164after. 165 166Additional considerations: 167* 4-space indents, no tabs, no trailing spaces, no last empty lines. If you 168 spot the code abusing these - thank you for fixing it. 169* Prefer keeping lines within 80 characters. Slight overflows are ok only 170 if that helps readability. 171* Please don't make "hypocritical structs" with protected or private members. 172 In general, `struct` is used to denote a plain-old-data structure, rather 173 than data+behaviour. If you need access control or are adding yet another 174 non-trivial (construction, assignment) member function to a `struct`, 175 just make it a `class` instead. 176* For newly created classes, keep to 177 [the rule of 3/5/0](http://en.cppreference.com/w/cpp/language/rule_of_three) - 178 make sure to read about the rule of zero if you haven't before, it's not 179 what you might think it is. 180* Qt containers are generally preferred to STL containers; however, there are 181 notable exceptions, and libQuotient already uses them: 182 * `std::array` and `std::deque` have no direct counterparts in Qt. 183 * Because of COW semantics, Qt containers cannot hold uncopyable classes. 184 Classes without a default constructor are a problem too. Examples of that 185 are `SyncRoomData` and `EventsArray<>`. Use STL containers for those but 186 see the next point and also consider if you can supply a reasonable 187 copy/default constructor. 188 * STL containers can be freely used in code internal to a translation unit 189 (i.e., in a certain .cpp file) _as long as that is not exposed in the API_. 190 It's ok to use, e.g., `std::vector` instead of `QVector` to tighten up 191 code where you don't need COW, or when dealing with uncopyable 192 data structures (see the previous point). However, exposing STL containers 193 in the API is not encouraged (except where absolutely necessary, e.g. we use 194 `std::deque` for a timeline). Exposing STL containers or iterators in API 195 intended for usage by QML code (e.g. in `Q_PROPERTY`) is unlikely to work 196 and therefore unlikely to be accepted into `master`. 197* Although `std::unique_ptr<>` gives slightly stronger guarantees, 198 `QScopedPointer<>` is better supported by Qt Creator's debugger UI and 199 therefore is preferred. 200* Use `QVector` instead of `QList` where possible - see the 201 [great article by Marc Mutz on Qt containers](https://marcmutz.wordpress.com/effective-qt/containers/) 202 for details. 203 204### API conventions 205 206Calls, data structures and other symbols not intended for use by clients 207should _not_ be exposed in (public) .h files, unless they are necessary 208to declare other public symbols. In particular, this involves private members 209(functions, typedefs, or variables) in public classes; use pimpl idiom to hide 210implementation details as much as possible. `_impl` namespace is reserved for 211definitions that should not be used by clients and are not covered by 212API guarantees. 213 214Note: As of now, all header files of libQuotient are considered public; this may change eventually. 215 216### Comments 217 218Whenever you add a new call to the library API that you expect to be used 219from client code, you must supply a proper doc-comment along with the call. 220Doxygen style is preferred; but JavaDoc is acceptable too. Some parts are 221not documented at all; adding doc-comments to them is highly encouraged. 222 223Doc-comments for summaries should be separate from those details. Either of 224the two following ways is fine, with considerable preference on the first one: 2251. Use `///` for the summary comment and `/*! ... */` for details. 2262. Use `\brief` (or `@brief`) for the summary, and follow with details after 227 an empty doc-comment line. You can use either of the delimiters in that case. 228 229In the code, the advice for commenting is as follows: 230* Don't restate what's happening in the code unless it's not really obvious. 231 We assume the readers to have at least some command of C++ and Qt. If your 232 code is not obvious, consider rewriting it for clarity. 233* Both C++ and Qt still come with their arcane features and dark corners, 234 and we don't want to limit anybody who'd feels they have a case for 235 variable templates, raw literals, or use `qAsConst` to avoid container 236 detachment. Use your experience to figure what might be less well-known to 237 readers and comment such cases (references to web pages, Quotient wiki etc. 238 are very much ok). 239* Make sure to document not so much "what" but more "why" certain code is done 240 the way it is. In the worst case, the logic of the code can be 241 reverse-engineered; you rarely can reverse-engineer the line of reasoning and 242 the pitfalls avoided. 243 244### Automated tests 245 246There's no testing framework as of now; either Catch or Qt Test or both will 247be used eventually. 248 249The `tests/` directory contains a command-line program, quotest, used for 250automated functional testing. Any significant addition to the library API 251should be accompanied by a respective test in quotest. To add a test you should: 252- Add a new test to the `TestSuite` class (technically, each test is a private 253 slot and there are two macros, `TEST_DECL()` and `TEST_IMPL()`, that conceal 254 passing the testing handle in `thisTest` variable to the test method). 255- Add test logic to the slot, using `FINISH_TEST` macro to assert the test 256 outcome and complete the test (`FINISH_TEST` contains `return`). ALL 257 (even failing) branches should conclude with a `FINISH_TEST` (or `FAIL_TEST` 258 that is a shortcut for a failing `FINISH_TEST`) invocation, unless you 259 intend to have a "DID NOT FINISH" message in the logs in certain conditions. 260 261The `TestManager` class sets up some basic test fixture to help you with testing; 262notably, the tests can rely on having an initialised `Room` object for the test 263room in `targetRoom` member variable. PRs to introduce a proper testing framework 264are very welcome (make sure to migrate tests from quotest though). Note that 265tests can go async, which is the biggest hurdle for Qt Test adoption. 266 267### Security, privacy, and performance 268 269Pay attention to security, and work *with* (not against) the usual security hardening mechanisms (however few in C++). 270 271`char *` and similar unchecked C-style read/write arrays are forbidden - use Qt containers or at the very least `std::array<>` instead. Where you see fit (usually with data structures), try to use smart pointers, especially `std::unique_ptr<>` or `QScopedPointer` instead of bare pointers. When dealing with `QObject`s, use the parent-child ownership semantics exercised by Qt (this is preferred to using smart pointers). Shared pointers are not used in the code so far; but if you find a particular use case where the strict semantic of unique pointers doesn't help and a shared pointer is necessary, feel free to step up with the working code and it will be considered for inclusion. 272 273Exercise the [principle of least privilege](https://en.wikipedia.org/wiki/Principle_of_least_privilege) where reasonable and appropriate. Prefer less-coupled cohesive code. 274 275Protect private information, in particular passwords and email addresses. Absolutely _don't_ spill around this information in logs - use `access_token` and similar opaque ids instead, and only display those in UI where needed. Do not forget about local access to data (in particular, be very careful when storing something in temporary files, let alone permanent configuration or state). Avoid mechanisms that could be used for tracking where possible (we do need to verify people are logged in but that's pretty much it), and ensure that third parties can't use interactions for tracking. Matrix protocols evolve towards decoupling the personally identifiable information from user activity entirely - follow this trend. 276 277We want the software to have decent performance for typical users. At the same time we keep libQuotient single-threaded as much as possible, to keep the code simple. That means being cautious about operation complexity (read about big-O notation if you need a kickstart on the topic). This especially refers to operations on the whole timeline and the list of users - each of these can have tens of thousands of elements so even operations with linear complexity, if heavy enough, can produce noticeable GUI freezing. When you don't see a way to reduce algorithmic complexity, embed occasional `processEvents()` invocations in heavy loops (see `Connection::saveState()` to get the idea). 278 279Having said that, there's always a trade-off between various attributes; in particular, readability and maintainability of the code is more important than squeezing every bit out of that clumsy algorithm. Beware of premature optimization and have profiling data around before going into some hardcore optimization. 280 281Speaking of profiling logs (see README.md on how to turn them on) - in order 282to reduce small timespan logging spam, there's a default limit of at least 283200 microseconds to log most operations with the PROFILER 284(aka quotient.profile.debug) logging category. You can override this 285limit by passing the new value (in microseconds) in PROFILER_LOG_USECS to 286the compiler. In the future, this parameter will be made changeable at runtime 287_if_ needed. 288 289### Generated C++ code for CS API 290The code in lib/csapi, lib/identity and lib/application-service, although 291it resides in Git, is actually generated from (a soft fork of) the official 292Matrix Swagger/OpenAPI definition files. Do not edit C++ files 293in these directories by hand! 294 295Now, if you're unhappy with something in there and want to improve the code, 296you have to understand the way these files are produced and setup 297some additional tooling. The shortest possible procedure resembling 298the below text can be found in .travis.yml (our Travis CI configuration 299actually regenerates those files upon every build). As described below, 300there is a handy build target for CMake; patches with a similar target 301for qmake are (you guessed it) very welcome. 302 303#### Why generate the code at all? 304Because before both original authors of libQuotient had to do monkey business of writing boilerplate code, with the same patterns, types etc., literally, for every single API endpoint, and one of the authors got fed up with it at some point in time. By then about 15 job classes were written; the entire API counts about 100 endpoints. Besides, the existing jobs had to be updated according to changes in CS API that have been, and will keep, coming. Other considerations can be found in [this talk about API description languages that briefly touches on GTAD](https://youtu.be/W5TmRozH-rg). 305 306#### Prerequisites for CS API code generation 3071. Get the source code of GTAD and its dependencies, e.g. using the command: 308 `git clone --recursive https://github.com/KitsuneRal/gtad.git` 3092. Build GTAD: in the source code directory, do `cmake . && cmake --build .` 310 (you might need to pass `-DCMAKE_PREFIX_PATH=<path to Qt>`, 311 similar to libQuotient itself). 3123. Get the Matrix CS API definitions that are included in the matrix-doc repo: 313 `git clone https://github.com/quotient-im/matrix-doc.git` 314 (quotient-im/matrix-doc is a fork that's known to produce working code; 315 you may want to use your own fork if you wish to alter something in the API). 3164. If you plan to submit a PR or just would like the generated code to be 317 formatted, you should either ensure you have clang-format (version 6 at least) 318 in your PATH or pass the _absolute_ path to it by adding 319 `-DCLANG_FORMAT=<absolute path>` to the CMake invocation below. 320 321#### Generating CS API contents 3221. Pass additional configuration to CMake when configuring libQuotient: 323 `-DMATRIX_DOC_PATH=<path you your matrix-doc repo> -DGTAD_PATH=<path to gtad binary (not the repo!)>`. 324 If everything's right, these two CMake variables will be mentioned in 325 CMake output and will trigger configuration of an additional build target, 326 see the next step. 3272. Generate the code: `cmake --build <your build dir> --target update-api`; 328 if you use CMake with GNU Make, you can just do `make update-api` instead. 329 Building this target will create (overwriting without warning) `.h` and `.cpp` 330 files in `lib/csapi`, `lib/identity`, `lib/application-service` for all 331 YAML files it can find in `matrix-doc/api/client-server` and other files 332 in `matrix-doc/api` these depend on. 3333. Re-run CMake so that the build system knows about new files, if there are any. 334 335#### Changing generated code 336See the more detailed description of what GTAD is and how it works in the documentation on GTAD in its source repo. Only parts specific for libQuotient are described here. 337 338GTAD uses the following three kinds of sources: 3391. OpenAPI files. Each file is treated as a separate source (because this is how GTAD works now). 3402. A configuration file, in our case it's lib/csapi/gtad.yaml - this one is common for the whole API. 3413. Source code template files: lib/csapi/{{base}}.*.mustache - also common. 342 343The mustache files have a templated (not in C++ sense) definition of a network 344job, deriving from BaseJob; each job class is prepended, if necessary, with 345data structure definitions used by this job. The look of those files is hideous 346for a newcomer; and the only known highlighter that can handle the combination 347of Mustache (originally a web templating language) and C++ is provided in CLion. 348To slightly simplify things some more or less generic constructs are defined 349in gtad.yaml (see its "mustache:" section). Adventurous souls that would like 350to figure what's going on in these files should speak up in the Quotient room - 351I (Kitsune) will be very glad to help you out. 352 353The types map in `gtad.yaml` is the central switchboard when it comes to matching OpenAPI types with C++ (and Qt) ones. It uses the following type attributes aside from pretty obvious "imports:": 354* `avoidCopy` - this attribute defines whether a const ref should be used instead of a value. For basic types like int this is obviously unnecessary; but compound types like `QVector` should rather be taken by reference when possible. 355* `moveOnly` - some types are not copyable at all and must be moved instead (an obvious example is anything "tainted" with a member of type `std::unique_ptr<>`). The template will use `T&&` instead of `T` or `const T&` to pass such types around. 356* `useOmittable` - wrap types that have no value with "null" semantics (i.e. number types and custom-defined data structures) into a special `Omittable<>` template defined in `converters.h` - a substitute for `std::optional` from C++17 (we're still at C++14 yet). 357* `omittedValue` - an alternative for `useOmittable`, just provide a value used for an omitted parameter. This is used for bool parameters which normally are considered false if omitted (or they have an explicit default value, passed in the "official" GTAD's `defaultValue` variable). 358* `initializer` - this is a _partial_ (see GTAD and Mustache documentation for explanations but basically it's a variable that is a Mustache template itself) that specifies how exactly a default value should be passed to the parameter. E.g., the default value for a `QString` parameter is enclosed into `QStringLiteral`. 359 360Instead of relying on the event structure definition in the OpenAPI files, `gtad.yaml` uses pointers to libQuotient's event structures: `EventPtr`, `RoomEventPtr` and `StateEventPtr`. Respectively, arrays of events, when encountered in OpenAPI definitions, are converted to `Events`, `RoomEvents` and `StateEvents` containers. When there's no way to figure the type from the definition, an opaque `QJsonObject` is used, leaving the conversion to the library and/or client code. 361 362## How to check proposed changes before submitting them 363 364Checking the code on at least one configuration is essential; if you only have 365a hasty fix that doesn't even compile, better make an issue and put a link to 366your commit into it (with an explanation what it is about and why). 367 368### Standard checks 369 370The following warnings configuration is applied with GCC and Clang when using CMake: 371`-W -Wall -Wextra -pedantic -Werror=return-type -Wno-unused-parameter -Wno-gnu-zero-variadic-macro-arguments` 372(the last one is to mute a warning triggered by Qt code for debug logging). 373We don't turn most of the warnings to errors but please treat them as such. 374If you use Qt Creator, the following line can be used with the Clang code model: 375`-Weverything -Werror=return-type -Wno-c++98-compat -Wno-c++98-compat-pedantic -Wno-unused-macros -Wno-newline-eof -Wno-exit-time-destructors -Wno-global-constructors -Wno-gnu-zero-variadic-macro-arguments -Wno-documentation -Wno-missing-prototypes -Wno-shadow-field-in-constructor -Wno-padded -Wno-weak-vtables -Wno-unknown-attributes -Wno-comma`. 376 377### Continuous Integration 378 379We use Travis CI to check buildability and smoke-testing on Linux (GCC, Clang) and MacOS (Clang), and AppVeyor CI to build on Windows (MSVC). Every PR will go through these, and you'll see the traffic lights from them on the PR page. If your PR fails on any platform double-check that it's not your code causing it - and fix it if it is. 380 381### clang-format 382 383We strongly recommend using clang-format or, even better, use an IDE that 384supports it. This will lay a tedious task of following the assumed 385code style from your shoulders over to your computer. 386 387### Other tools 388 389Recent versions of Qt Creator and CLion can automatically run your code through 390clang-tidy. The following list of clang-tidy checks slows Qt Creator analysis 391quite considerably but gives a good insight without too many false positives: 392`-*,bugprone-argument-comment,bugprone-assert-side-effect,bugprone-bool-pointer-implicit-conversion,bugprone-copy-constructor-init,bugprone-dangling-handle,bugprone-fold-init-type,bugprone-forward-declaration-namespace,bugprone-inaccurate-erase,bugprone-integer-division,bugprone-move-forwarding-reference,bugprone-string-constructor,bugprone-undefined-memory-manipulation,bugprone-use-after-move,bugprone-virtual-near-miss,cert-dcl03-c,cert-dcl21-cpp,cert-dcl50-cpp,cert-dcl54-cpp,cert-dcl58-cpp,cert-env33-c,cert-err09-cpp,cert-err34-c,cert-err52-cpp,cert-err60-cpp,cert-err61-cpp,cert-fio38-c,cert-flp30-c,cert-msc30-c,cert-msc50-cpp,cert-oop11-cpp,cppcoreguidelines-c-copy-assignment-signature,cppcoreguidelines-pro-type-cstyle-cast,cppcoreguidelines-slicing,hicpp-deprecated-headers,hicpp-invalid-access-moved,hicpp-member-init,hicpp-move-const-arg,hicpp-named-parameter,hicpp-new-delete-operators,hicpp-static-assert,hicpp-undelegated-constructor,hicpp-use-*,misc-misplaced-const,misc-new-delete-overloads,misc-non-copyable-objects,misc-redundant-expression,misc-static-assert,misc-throw-by-value-catch-by-reference,misc-unconventional-assign-operator,misc-uniqueptr-reset-release,misc-unused-*,modernize-loop-convert,modernize-pass-by-value,modernize-return-braced-init-list,modernize-shrink-to-fit,modernize-unary-static-assert,modernize-use-*,performance-faster-string-find,performance-for-range-copy,performance-implicit-conversion-in-loop,performance-inefficient-*,performance-move-*,performance-type-promotion-in-math-fn,performance-unnecessary-*,readability-delete-null-pointer,readability-else-after-return,readability-inconsistent-declaration-parameter-name,readability-misleading-indentation,readability-redundant-*,readability-simplify-boolean-expr,readability-static-definition-in-anonymous-namespace,readability-uniqueptr-delete-release`. 393 394Qt Creator, in addition, knows about clazy, an even deeper Qt-aware static 395analysis tool. Even level 1 clazy eats away CPU but produces some very relevant 396notices that are easy to overlook otherwise, such as possible unintended copying 397of a Qt container, or unguarded null pointers. You can use this time to time 398(see Analyze menu in Qt Creator) instead of hogging your machine with 399deep analysis as you type. 400 401## Git commit messages 402 403When writing git commit messages, try to follow the guidelines in 404[How to Write a Git Commit Message](https://chris.beams.io/posts/git-commit/): 405 4061. Separate subject from body with a blank line 4072. Be reasonable on the subject line length, because this is what we see in commit logs. Try to fit in 50 characters whenever possible. 4083. Capitalize the subject line 4094. Do not end the subject line with a period 4105. Use the imperative mood in the subject line (*command* form) 411 (we don't always practice this ourselves but let's try). 4126. Use the body to explain what and why vs. how 413 (git tracks how it was changed in detail, don't repeat that). Sometimes a quick overview of "how" is acceptable if a commit is huge - but maybe split a commit into smaller ones, to begin with? 414 415## Reuse (libraries, frameworks, etc.) 416 417C++ is unfortunately not very coherent about SDK/package management, and we try to keep building the library as easy as possible. Because of that we are very conservative about adding dependencies to libQuotient. That relates to additional Qt components and even more to other libraries. Fortunately, even the Qt components now in use (Qt Core and Network) are very feature-rich and provide plenty of ready-made stuff. 418 419Regardless of the above paragraph (and as mentioned earlier in the text), we're now looking at possible options for futures and automated testing, so PRs onboarding those will be considered with much gratitude. 420 421Some cases need additional explanation: 422* Before rolling out your own super-optimised container or algorithm written 423 from scratch, take a good long look through documentation on Qt and 424 C++ standard library. Please try to reuse the existing facilities 425 as much as possible. 426* You should have a good reason (or better several ones) to add a component 427 from KDE Frameworks. We don't rule this out and there's no prejudice against 428 KDE; it just so happened that KDE Frameworks is one of most obvious 429 reuse candidates but so far none of these components survived 430 as libQuotient deps. So we are cautious. Extra notice to KDE folks: 431 I'll be happy if an addon library on top of libQuotient is made using 432 KDE facilities, and I'm willing to take part in its evolution; but please 433 also respect LXDE people who normally don't have KDE frameworks installed. 434* Never forget that libQuotient is aimed to be a non-visual library; 435 QtGui in dependencies is only driven by (entirely offscreen) dealing with 436 QImages. While there's a bunch of visual code (in C++ and QML) shared 437 between Quotient-enabled _applications_, this is likely to end up 438 in a separate (Quotient-enabled) library, rather than libQuotient itself. 439 440## Attribution 441 442This text is based on CONTRIBUTING.md from CII Best Practices Badge project, which is a collective work of its contributors (many thanks!). The text itself is licensed under CC-BY-4.0. 443