Using C++ in Mozilla code

C++ language features

Mozilla code only uses a subset of C++. Runtime type information (RTTI) is disabled, as it tends to cause a very large increase in codesize. This means that dynamic_cast, typeid() and <typeinfo> cannot be used in Mozilla code. Also disabled are exceptions; do not use try/catch or throw any exceptions. Libraries that throw exceptions may be used if you are willing to have the throw instead be treated as an abort.

On the side of extending C++, we compile with -fno-strict-aliasing. This means that when reinterpreting a pointer as a differently-typed pointer, you don't need to adhere to the "effective type" (of the pointee) rule from the standard (aka. "the strict aliasing rule") when dereferencing the reinterpreted pointer. You still need make sure that you don't violate alignment requirements and need to make sure that the data at the memory location pointed to forms a valid value when interpreted according to the type of the pointer when dereferencing the pointer for reading. Likewise, if you write by dereferencing the reinterpreted pointer and the originally-typed pointer might still be dereferenced for reading, you need to make sure that the values you write are valid according to the original type. This value validity issue is moot for e.g. primitive integers for which all bit patterns of their size are valid values.

  • As of Mozilla 59, C++14 mode is required to build Mozilla.
  • As of Mozilla 67, MSVC can no longer be used to build Mozilla.
  • As of Mozilla 73, C++17 mode is required to build Mozilla.

This means that C++17 can be used where supported on all platforms. The list of acceptable features is given below:

GCC Clang
Current minimal requirement 7.1 5.0
Feature GCC Clang Can be used in code
type_t && 4.3 2.9 Yes (see notes)
ref qualifiers on methods 4.8.1 2.9 Yes
default member initializers (except for bit-fields) 4.7 3.0 Yes
default member initializers (for bit-fields) 8 6 No
variadic templates 4.3 2.9 Yes
Initializer lists 4.4 3.1 Yes
static_assert 4.3 2.9 Yes
auto 4.4 2.9 Yes
lambdas 4.5 3.1 Yes
decltype 4.3 2.9 Yes
Foo<Bar<T>> 4.3 2.9 Yes
auto func() -> int 4.4 3.1 Yes
Templated aliasing 4.7 3.0 Yes
nullptr 4.6 3.0 Yes
enum foo : int16_t {}; 4.4 2.9 Yes
enum class foo {}; 4.4 2.9 Yes
enum foo; 4.6 3.1 Yes
[[attributes]] 4.8 3.3 No (see notes)
constexpr 4.6 3.1 Yes
alignas 4.8 3.3 Yes
alignof 4.8 3.3 Yes, but see notes ; only clang 3.6 claims __has_feature(cxx_alignof)
Delegated constructors 4.7 3.0 Yes
Inherited constructors 4.8 3.3 Yes
explicit operator bool() 4.5 3.0 Yes
char16_t/u"string" 4.4 3.0 Yes
R"(string)" 4.5 3.0 Yes
operator""() 4.7 3.1 Yes
=delete 4.4 2.9 Yes
=default 4.4 3.0 Yes
unrestricted unions 4.6 3.1 Yes
for (auto x : vec) 4.6 3.0 Yes (be careful about the type of the iterator)
override/final 4.7 3.0 Yes
thread_local 4.8 3.3 No (see notes)
function template default arguments 4.3 2.9 Yes
local structs as template parameters 4.5 2.9 Yes
extended friend declarations 4.7 2.9 Yes
0b100 (C++14) 4.9 2.9 Yes
Tweaks to some C++ contextual conversions (C++14) 4.9 3.4 Yes
Return type deduction (C++14) 4.9 3.4 Yes (but only in template code when you would have used decltype(complex-expression))
Generic lambdas (C++14) 4.9 3.4 Yes
Initialized lambda captures (C++14) 4.9 3.4 Yes
Digit separator (C++14) 4.9 3.4 Yes
Variable templates (C++14) 5.0 3.4 Yes
Relaxed constexpr (C++14) 5.0 3.4 Yes
Aggregate member initialization (C++14) 5.0 3.3 Yes
Clarifying memory allocation (C++14) 5.0 3.4 Yes
[[deprecated]] attribute (C++14) 4.9 3.4 No (see notes)
Sized deallocation (C++14) 5.0 3.4 No (see notes)
Concepts (Concepts TS) 6.0 No
Inline variables (C++17) 7.0 3.9 No (clang 5 has bugs with inline variables)
constexpr_if (C++17) 7.0 3.9 Yes
constexpr lambdas (C++17) No
Structured bindings (C++17) 7.0 4.0 Yes
[[fallthrough]], [[maybe_unused]], [[nodiscard]] (C++17) 7.0 3.9 Yes
Aligned allocation/deallocation (C++17) 7.0 4.0 No (see notes)
#pragma once 3.4 Yes Not until we normalize headers
Source code information capture 8.0 No

Sources

GCC: https://gcc.gnu.org/projects/cxx-status.html
Clang: https://clang.llvm.org/cxx_status.html

Notes

rvalue references: Implicit move method generation cannot be used.

Attributes: Several common attributes are defined in mozilla/Attributes.h or nscore.h.

Alignment: Some alignment utilities are defined in mozilla/Alignment.h. /!\ MOZ_ALIGNOF and alignof don't have the same semantics. Be careful of what you expect from them.

[[deprecated]]: If we have deprecated code, we should be removing it rather than marking it as such. Marking things as [[deprecated]] also means the compiler will warn if you use the deprecated API, which turns into a fatal error in our automation builds, which is not helpful.

Sized deallocation: Our compilers all support this (custom flags are required for GCC and Clang), but turning it on breaks some classes' operator new methods, and some work would need to be done to make it an efficiency win with our custom memory allocator.

Aligned allocation/deallocation: Our custom memory allocator doesn't have support for these functions.

Thread locals: thread_local is not supported on Android.

C++ and Mozilla standard libraries

The Mozilla codebase contains within it several subprojects which follow different rules for which libraries can and can't be used it. The rules listed here apply to normal platform code, and assume unrestricted usability of MFBT or XPCOM APIs.

The rest of this section is a draft for expository and exploratory purposes. Do not trust the information listed here.

What follows is a list of standard library components provided by Mozilla or the C++ standard. If an API is not listed here, then it is not permissible to use it in Mozilla code. Deprecated APIs are not listed here. In general, prefer Mozilla variants of data structures to standard C++ ones, even when permitted to use the latter, since Mozilla variants tend to have features not found in the standard library (e.g., memory size tracking) or have more controllable performance characteristics.

A list of approved standard library headers is maintained in config/stl-headers.mozbuild.

Data structures

Name Header STL equivalent Notes
nsAutoTArray nsTArray.h Like nsTArray, but will store a small amount as stack storage
nsAutoTObserverArray nsTObserverArray.h Like nsTObserverArray, but will store a small amount as stack storage
mozilla::BloomFilter mozilla/BloomFilter.h Probabilistic set membership (see Wikipedia)
nsClassHashtable nsClassHashtable.h Adaptation of nsTHashtable, see XPCOM hashtable guide
nsCOMArray nsCOMArray.h Like nsTArray<nsCOMPtr<T>>
nsDataHashtable nsClassHashtable.h std::unordered_map Adaptation of nsTHashtable, see XPCOM hashtable guide
nsDeque nsDeque.h std::deque<void *>
mozilla::EnumSet mozilla/EnumSet.h Like std::set, but for enum classes.
mozilla::Hash{Map,Set} mozilla/HashTable.h std::unordered_{map,set} A general purpose hash map and hash set.
nsInterfaceHashtable nsInterfaceHashtable.h std::unordered_map Adaptation of nsTHashtable, see XPCOM hashtable guide
nsJSThingHashtable nsJSThingHashtable.h Adaptation of nsTHashtable, see XPCOM hashtable guide
mozilla::LinkedList mozilla/LinkedList.h std::list Doubly-linked list
nsRefPtrHashtable nsRefPtrHashtable.h std::unordered_map Adaptation of nsTHashtable, see XPCOM hashtable guide
mozilla::SegmentedVector mozilla/SegmentedVector.h std::deque w/o O(1) pop_front Doubly-linked list of vector elements
mozilla::SplayTree mozilla/SplayTree.h Quick access to recently-accessed elements (see Wikipedia)
nsTArray nsTArray.h std::vector
nsTHashtable nsTHashtable.h std::unordered_{map,set} See XPCOM hashtable guide, you probably want a subclass
nsTObserverArray nsTObserverArray.h Like nsTArray, but iteration is stable even through mutation
nsTPriorityQueue nsTPriorityQueue.h std::priority_queue Unlike the STL class, not a container adapter
mozilla::Vector mozilla/Vector.h std::vector
mozilla::Buffer mozilla/Buffer.h Unlike Array, has a run-time variable length. Unlike Vector, does not have capacity and growth mechanism. Unlike Span, owns its buffer.

Safety utilities

Name Header STL equivalent Notes
mozilla::Array mfbt/Array.h safe array index
mozilla::AssertedCast mfbt/Casting.h casts
mozilla::CheckedInt mfbt/CheckedInt.h avoids overflow
nsCOMPtr xpcom/base/nsCOMPtr.h std::shared_ptr
mozilla::EnumeratedArray mfbt/EnumeratedArray.h mozilla::Array
mozilla::Maybe mfbt/Maybe.h std::optional
mozilla::RangedPtr mfbt/RangedPtr.h like mozilla::Span but with two pointers instead of pointer and length
mozilla::RefPtr mfbt/RefPtr.h std::shared_ptr
mozilla::Span mozilla/Span.h gsl::span, absl::Span, std::string_view, std::u16string_view Rust's slice concept for C++ (without borrow checking)
StaticRefPtr xpcom/base/StaticPtr.h nsRefPtr w/o static constructor
mozilla::UniquePtr mfbt/UniquePtr.h std::unique_ptr
mozilla::WeakPtr mfbt/WeakPtr.h std::weak_ptr
nsWeakPtr xpcom/base/nsWeakPtr.h std::weak_ptr

Strings

See the Mozilla internal string guide for usage of nsAString (our copy-on-write replacement for std::u16string) and nsACString (our copy-on-write replacement for std::string).

Be sure not to introduce further uses of std::wstring, which is not portable! (Some uses exist in the IPC code.)

Algorithms

mozilla::BinarySearch mfbt/BinarySearch.h
mozilla::BitwiseCast mfbt/Casting.h (strict aliasing-safe cast)
mozilla/MathAlgorithms.h (rotate, ctlz, popcount, gcd, abs, lcm)
mozilla::RollingMean mfbt/RollingMean.h ()

Concurrency

Name Header STL/boost equivalent Notes
mozilla::Atomic mfbt/Atomic.h std::atomic
mozilla::CondVar xpcom/threads/CondVar.h std::condition_variable
mozilla::DataMutex xpcom/threads/DataMutex.h boost::synchronized_value
mozilla::Monitor xpcom/threads/Monitor.h
mozilla::Mutex xpcom/threads/Mutex.h std::mutex
mozilla::ReentrantMonitor xpcom/threads/ReentrantMonitor.h
mozilla::StaticMutex xpcom/base/StaticMutex.h std::mutex Mutex that can (and in fact, must) be used as a global/static variable.

Miscellaneous

Name Header STL/boost equivalent Notes
mozilla::AlignedStorage mfbt/Alignment.h std::aligned_storage
mozilla::MaybeOneOf mfbt/MaybeOneOf.h std::optional<std::variant<T1, T2>> ~mozilla::Maybe<union {T1, T2}>
mozilla::Pair mfbt/Pair.h std::tuple<T1, T2> minimal space!
mozilla::TimeStamp xpcom/ds/TimeStamp.h std::chrono::time_point
mozilla/TypeTraits.h <type_traits>
mozilla/PodOperations.h C++ versions of memset, memcpy, etc.
mozilla/ArrayUtils.h
mozilla/Compression.h
mozilla/Endian.h
mozilla/FloatingPoint.h
mozilla/HashFunctions.h std::hash
mozilla/Move.h std::move, std::swap, std::forward

Mozilla data structures and standard C++ ranges and iterators

Some Mozilla-defined data structures provide STL-style iterators and are usable in range-based for loops as well as STL algorithms.

Currently, these include:

Name Header Bug(s) Iterator category Notes
nsTArray xpcom/ds/nsTArray.h 1126552 Random-access

Also reverse-iterable.
Also supports remove-erase pattern via RemoveElementsAt method.
Also supports back-inserting output iterators via MakeBackInserter function.

nsBaseHashtable and subclasses:
nsClassHashtable
nsDataHashtable
nsInterfaceHashtable
nsJSThingHashtable
nsRefPtrHashtable

xpcom/ds/nsBaseHashtable.h
xpcom/ds/nsClassHashtable.h
xpcom/ds/nsDataHashtable.h
xpcom/ds/nsInterfaceHashtable.h
xpcom/ds/nsJSThingHashtable.h
xpcom/ds/nsRefPtrHashtable.h

1575479 Forward
nsCOMArray xpcom/ds/nsCOMArray.h 1342303 Random-access Also reverse-iterable.

Array
EnumerationArray
RangedArray

mfbt/Array.h
mfbt/EnumerationArray.h
mfbt/RangedArray.h

1216041 Random-access Also reverse-iterable.
Buffer mfbt/Buffer.h 1512155 Random-access Also reverse-iterable.
DoublyLinkedList mfbt/DoublyLinkedList.h 1277725 Forward
EnumeratedRange mfbt/EnumeratedRange.h 1142999 Missing Also reverse-iterable.
IntegerRange mfbt/IntegerRange.h 1126701 Missing Also reverse-iterable.
SmallPointerArray mfbt/SmallPointerArray.h 1331718 Random-access
Span mfbt/Span.h 1295611 Random-access Also reverse-iterable.

Note that if the iterator category is stated as "missing", the type is probably only usable in range-based for. This is most likely just an omission, which could be easily fixed.

Useful in this context are also the class template IteratorRange (which can be used to construct a range from any pair of iterators) and function template Reversed (which can be used to reverse any range), both defined in mfbt/ReverseIterator.h

Further C++ rules

Don't use static constructors

(You probably shouldn't be using global variables to begin with. Quite apart from the weighty software-engineering arguments against them, globals affect startup time! But sometimes we have to do ugly things.)

Non-portable example:

FooBarClass static_object(87, 92);

void
bar()
{
  if (static_object.count > 15) {
     ...
  }
}

Once upon a time, there were compiler bugs that could result in constructors not being called for global objects. Those bugs are probably long gone by now, but even with the feature working correctly, there are so many problems with correctly ordering C++ constructors that it's easier to just have an init function:

static FooBarClass* static_object;

FooBarClass*
getStaticObject()
{
  if (!static_object)
    static_object =
      new FooBarClass(87, 92);
  return static_object;
}

void
bar()
{
  if (getStaticObject()->count > 15) {
    ...
  }
}

Don't use exceptions

See the introduction to the "C++ language features" section at the start of this document.

Don't use Run-time Type Information

See the introduction to the "C++ language features" section at the start of this document.

If you need runtime typing, you can achieve a similar result by adding a classOf() virtual member function to the base class of your hierarchy and overriding that member function in each subclass. If classOf() returns a unique value for each class in the hierarchy, you'll be able to do type comparisons at runtime.

Don't use the C++ standard library (including iostream and locale)

See the section "C++ and Mozilla standard libraries".

Use C++ lambdas, but with care

C++ lambdas are supported across all our compilers now.  Rejoice!  We recommend explicitly listing out the variables that you capture in the lambda, both for documentation purposes, and to double-check that you're only capturing what you expect to capture.

Use namespaces

Namespaces may be used according to the style guidelines in Mozilla Coding Style Guide.

Don't mix varargs and inlines

What? Why are you using varargs to begin with?! Stop that at once!

Make header files compatible with C and C++

Non-portable example:

/*oldCheader.h*/
int existingCfunction(char*);
int anotherExistingCfunction(char*);

/* oldCfile.c */
#include "oldCheader.h"
...

// new file.cpp
extern "C" {
#include "oldCheader.h"
};
...

If you make new header files with exposed C interfaces, make the header files work correctly when they are included by both C and C++ files.

(If you need to include a C header in new C++ files, that should just work. If not, it's the C header maintainer's fault, so fix the header if you can, and if not, whatever hack you come up with will probably be fine.)

Portable example:

/* oldCheader.h*/
PR_BEGIN_EXTERN_C
int existingCfunction(char*);
int anotherExistingCfunction(char*);
PR_END_EXTERN_C

/* oldCfile.c */
#include "oldCheader.h"
...

// new file.cpp
#include "oldCheader.h"
...

There are number of reasons for doing this, other than just good style. For one thing, you are making life easier for everyone else, doing the work in one common place (the header file) instead of all the C++ files that include it. Also, by making the C header safe for C++, you document that "hey, this file is now being included in C++". That's a good thing. You also avoid a big portability nightmare that is nasty to fix...

Use override on subclass virtual member functions

The override keyword is supported in C++11 and in all our supported compilers, and it catches bugs.

Always declare a copy constructor and assignment operator

Many classes shouldn't be copied or assigned. If you're writing one of these, the way to enforce your policy is to declare a deleted copy constructor as private and not supply a definition. While you're at it, do the same for the assignment operator used for assignment of objects of the same class. Example:

class Foo {
  ...
  private:
    Foo(const Foo& x) = delete;
    Foo& operator=(const Foo& x) = delete;
};

Any code that implicitly calls the copy constructor will hit a compile-time error. That way nothing happens in the dark. When a user's code won't compile, they'll see that they were passing by value, when they meant to pass by reference (oops).

Be careful of overloaded methods with like signatures

It's best to avoid overloading methods when the type signature of the methods differs only by one "abstract" type (e.g. PR_Int32 or int32). What you will find as you move that code to different platforms, is suddenly on the Foo2000 compiler your overloaded methods will have the same type-signature.

Type scalar constants to avoid unexpected ambiguities

Non-portable code:

class FooClass {
  // having such similar signatures
  // is a bad idea in the first place.
  void doit(long);
  void doit(short);
};

void
B::foo(FooClass* xyz)
{
  xyz->doit(45);
}

Be sure to type your scalar constants, e.g., uint32_t(10) or 10L. Otherwise, you can produce ambiguous function calls which potentially could resolve to multiple methods, particularly if you haven't followed (2) above. Not all of the compilers will flag ambiguous method calls.

Portable code:

class FooClass {
  // having such similar signatures
  // is a bad idea in the first place.
  void doit(long);
  void doit(short);
};

void
B::foo(FooClass* xyz)
{
  xyz->doit(45L);
}

Use nsCOMPtr in XPCOM code

See the nsCOMPtr User Manual for usage details.

Don't use identifiers that start with an underscore

This rule occasionally surprises people who've been hacking C++ for decades. But it comes directly from the C++ standard!

According to the C++ Standard, 17.4.3.1.2 Global Names [lib.global.names], paragraph 1:

Certain sets of names and function signatures are always reserved to the implementation:

  • Each name that contains a double underscore (__) or begins with an underscore followed by an uppercase letter (2.11) is reserved to the implementation for any use.
  • Each name that begins with an underscore is reserved to the implementation for use as a name in the global namespace.

Stuff that is good to do for C or C++

Avoid conditional #includes when possible

Don't write an #include inside an #ifdef if you could instead put it outside. Unconditional includes are better because they make the compilation more similar across all platforms and configurations, so you're less likely to cause stupid compiler errors on someone else's favorite platform that you never use.

Bad code example:

#ifdef MOZ_ENABLE_JPEG_FOUR_BILLION
#include <stdlib.h>   // <--- don't do this
#include "jpeg4e9.h"  // <--- only do this if the header really might not be there
#endif

Of course when you're including different system files for different machines, you don't have much choice. That's different.

Every .cpp source file should have a unique name

Every object file linked into libxul needs to have a unique name. Avoid generic names like nsModule.cpp and instead use nsPlacesModule.cpp.

Turn on warnings for your compiler, and then write warning free code

What generates a warning on one platform will generate errors on another. Turn warnings on. Write warning-free code. It's good for you. Treat warnings as errors by adding ac_add_options --enable-warnings-as-errors to your mozconfig file.

Use the same type for all bitfields in a struct or class

Some compilers do not pack the bits when different bitfields are given different types. For example, the following struct might have a size of 8 bytes, even though it would fit in 1:

struct {
  char ch : 1;
  int i : 1;
};

Don't use an enum type for a bitfield

The classic example of this is using PRBool for a boolean bitfield. Don't do that. PRBool is a signed integer type, so the bitfield's value when set will be -1 instead of +1, which---I know, crazy, right? The things C++ hackers used to have to put up with...

You shouldn't be using PRBool anyway. Use bool. Bitfields of type bool are fine.

Enums are signed on some platforms (in some configurations) and unsigned on others and therefore unsuitable for writing portable code when every bit counts, even if they happen to work on your system.