solvespace/CONTRIBUTING.md

10 KiB

Contributing to SolveSpace

Contributing bug reports

Bug reports are always welcome! When reporting a bug, please include the following:

  • The version of SolveSpace (use Help → About...);
  • The operating system;
  • The save file that reproduces the incorrect behavior, or, if trivial or impossible, instructions for reproducing it.

GitHub does not allow attaching *.slvs files, but it does allow attaching *.zip files, so any savefiles should first be archived.

Contributing translations

To contribute a translation, not a lot is necessary—at a minimum, you need to be able to edit .po files with a tool such as poedit. Once you have such a tool installed, take res/messages.pot and start translating!

However, if you want to see your translation in action, a little more work is necessary. First, you need to be able to build SolveSpace; see README. After that:

  • Copy res/messages.pot to res/locales/xx_YY.po, where xx is an ISO 639-1 country code, and YY is an ISO 3166-1 language code.
  • Add a line xx-YY,LCID,Name to res/locales.txt, where xx-YY have the same meaning as above, LCID is a Windows Language Code Identifier (MS-LCID has a complete list), and Name is the full name of your locale in your language.
  • Add locales/xx_YY.po in res/CMakeLists.txt—search for locales/en_US.po to see where it should be added.

You're done! Recompile SolveSpace and you should be able to select your translation via Help → Language.

Contributing code

SolveSpace is written in C++, and currently targets all compilers compliant with C++11. This includes GCC 5 and later, Clang 3.3 and later, and Visual Studio 12 (2013) and later.

High-level conventions

Portability

SolveSpace aims to consist of two general parts: a fully portable core, and platform-specific UI and support code. Anything outside of src/platform/ should only use standard C++11, and rely on src/platform/unixutil.cpp and src/platform/w32util.cpp to interact with the OS where this cannot be done through the C++11 standard library.

Libraries

SolveSpace primarily relies on the C++11 STL. STL has well-known drawbacks, but is also widely supported, used, and understood. SolveSpace also includes a fair amount of use of bespoke containers List and IdList; these provide STL iterators, and can be used when convenient, such as when reusing other code.

One notable departure here is the STL I/O threads. SolveSpace does not use STL I/O threads for two reasons: (i) the interface is borderline unusable, and (ii) on Windows it is not possible to open files with Unicode paths through STL.

When using external libraries (other than to access platform features), the libraries should satisfy the following conditions:

  • Portable, and preferably not interacting with the platform at all;
  • Can be included as a CMake subproject, to facilitate Windows, Android, etc. builds;
  • Use a license less restrictive than GPL (BSD/MIT, Apache2, MPL, etc.)

String encoding

Internally, SolveSpace exclusively stores and uses UTF-8 for all purposes; any std::string may be assumed to be encoded in UTF-8. On Windows, UTF-8 strings are converted to and from wide strings at the boundary; see UTF-8 Everywhere for details.

String formatting

For string formatting, a wrapper around sprintf, ssprintf, is used. A notable pitfall when using it is trying to pass an std::string argument without first converting it to a C string with .c_str().

Filesystem access

For filesystem access, the C standard library is used. The ssfopen and ssremove wrappers are provided that accept UTF-8 encoded paths.

Assertions

To ensure that internal invariants hold, the ssassert function is used, e.g. ssassert(!isFoo, "Unexpected foo condition");. Unlike the standard assert function, the ssassert function is always enabled, even in release builds. It is more valuable to discover a bug through a crash than to silently generate incorrect results, and crashes do not result in losing more than a few minutes of work thanks to the autosave feature.

Use of C++ features

The conventions described in this section should be used for all new code, but there is a lot of existing code in SolveSpace that does not use them. This is fine; don't touch it if it works, but if you need to modify it anyway, might as well modernize it.

Exceptions

Exceptions are not used primarily because SolveSpace's testsuite uses measurement of branch coverage, important for the critical parts such as the geometric kernel. Every function call with exceptions enabled introduces a branch, making branch coverage measurement useless.

Operator overloading

Operator overloading is not used primarily for historical reasons. Instead, method such as Plus are used.

Member visibility

Member visibility is not used for implementation hiding. Every member field and function is public.

Constructors

Constructors are not used for initialization, chiefly because indicating an error in a constructor would require throwing an exception, nor does it use constructors for blanket zero-initialization because of the performance impact of doing this for common POD classes like Vector.

Instances can be zero-initialized using the aggregate-initialization syntax, e.g. Foo foo = {};. This zero-initializes the POD members and default-initializes the non-POD members, generally being an equivalent of memset(&foo, 0, sizeof(foo)); but compatible with STL containers.

Input- and output-arguments

Functions accepting an input argument take it either by-value (Vector v) or by-const-reference (const Vector &v). Generally, passing by-value is safer as the value cannot be aliased by something else, but passing by-const-reference is faster, as a copy is eliminated. Small values should always be passed by-value, and otherwise functions that do not capture pointers into their arguments should take them by-const-reference. Use your judgement.

Functions accepting an output argument always take it by-pointer (Vector *v). This makes it immediately visible at the call site as it is seen that the address is taken. Arguments are never passed by-reference, except when needed for interoperability with STL, etc.

Iteration

foreach-style iteration is preferred for both STL and List/IdList containers as it indicates intent clearly, as opposed to for-style.

Const correctness

Functions that do not mutate this should be marked as const; when iterating a collection without mutating any of its elements, for(const Foo &elem : collection) is preferred to indicate the intent.

Coding style

Code is formatted by the following rules:

  • Code is indented using 4 spaces, with no trailing spaces, and lines are wrapped at 100 columns;
  • Braces are placed at the end of the line with the declaration or control flow statement;
  • Braces are used with every control flow statement, even if there is only one statement in the body;
  • There is no space after control flow keywords (if, while, etc.);
  • Identifiers are formatted in camel case; variables start with a lowercase letter (exampleVariable) and functions start with an uppercase letter (ExampleFunction).

For example:

std::string SolveSpace::Dirname(std::string filename) {
    int slash = filename.rfind(PATH_SEP);
    if(slash >= 0) {
        return filename.substr(0, slash);
    }

    return "";
}

Debugging code

SolveSpace releases are throughly tested but sometimes they contain crash bugs anyway. The reason for such crashes can be determined only if the executable was built with debug information.

Debugging a released version

The Linux distributions usually include separate debug information packages. On a Debian derivative (e.g. Ubuntu), these can be installed with:

apt-get install solvespace-dbg

The macOS releases include the debug information, and no further action is needed.

The Windows releases include the debug information on the GitHub release downloads page.

Debugging a custom build

If you are building SolveSpace yourself on macOS, use the XCode CMake generator, then open the project in XCode as usual, select the Debug build scheme, and build the project:

cd build
cmake .. -G Xcode [other cmake args...]

If you are building SolveSpace yourself on any Unix-like platform, configure or re-configure SolveSpace to produce a debug build, and then build it:

cd build
cmake .. -DCMAKE_BUILD_TYPE=Debug [other cmake args...]
make

If you are building SolveSpace yourself using the Visual Studio IDE, select Debug from the Solution Configurations list box on the toolbar, and build the solution.

Debugging with gdb

gdb is a debugger that is mostly used on Linux. First, run SolveSpace under debugging:

gdb [path to solvespace executable]
(gdb) run

Then, reproduce the crash. After the crash, attach the output in the console, as well as output of the following gdb commands to a bug report:

(gdb) backtrace
(gdb) info locals

If the crash is not easy to reproduce, please generate a core file, which you can use to resume the debugging session later, and provide any other information that is requested:

(gdb) generate-core-file

This will generate a large file called like core.1234 in the current directory; it can be later re-loaded using gdb --core core.1234.

Debugging with lldb

lldb is a debugger that is mostly used on macOS. First, run SolveSpace under debugging:

lldb [path to solvespace executable]
(lldb) run

Then, reproduce the crash. After the crash, attach the output in the console, as well as output of the following gdb commands to a bug report:

(lldb) backtrace all
(lldb) frame variable

If the crash is not easy to reproduce, please generate a core file, which you can use to resume the debugging session later, and provide any other information that is requested:

(lldb) process save-core "core"

This will generate a large file called core in the current directory; it can be later re-loaded using lldb -c core.

There are several environment variables available that make crashes earlier and errors more informative. Before running SolveSpace, run the following commands in your shell:

export G_DEBUG=fatal_warnings
export LIBGL_DEBUG=1
export MESA_DEBUG=1