Coding Style

From musl libc wiki
Jump to: navigation, search

At present, musl does not have a strict coding style policy. The best short description of the coding style used in musl is "very similar to the Linux kernel style". The style guidelines which follow are descriptive rather than prescriptive; they document what's currently in use in musl, and hopefully give a feel for how new code should look.

Indention

The indention style in musl distinguishes between "indention" and "alignment". Indention is leading whitespace that serves as a visual indicator of the block structure of the code. Alignment is initial or internal whitespace used to line up content on consecutive lines (e.g. a sequence of macro definitions or typedefs or array initializers) or continuations of a line or statement onto additional physical lines.

Indention is purely tabs, no spaces. Alignment is purely spaces, no tabs. A line may have initial tabs followed by initial spaces if it's subject to both indention and alignment. When a statement is split onto multiple lines, either indention (1 additional level) or alignment (to align with some meaningful location in the first line) may be used.

This style permits the code to be viewed correctly with any tab width setting.

Column width

Code should aim to fit within 80 columns, with the default tab setting of 8 spaces. Breaking this rule is encouraged if the overflow is minimal and the content that would be cut off when viewing at 80 columns is mostly uninteresting or obvious.

Spacing

Spaces are not used after casts or before the opening parenthesis of a function call. Aside from these rules, spacing is one of the most flexible parts of the coding style. It is desirable to adjust spacing to avoid breaking lines or going over 80 columns, and to reflect grouping and operator precedence.

Redundant type specifiers

The shortest possible name of a type should be used, especially in public headers. Redundant type specifiers should not be used. This means "short" is preferred over "short int" or "signed short int".

Goto

Goto should be used sparingly, mainly to unify handling of errors that occur at different points with common error handling code. In some cases it's also used for adding "retry" logic to a function non-invasively, where it's undesirable for the structure of the function to be shaped around the possibility of having to retry. In a few cases, it's also used for starting the first iteration of a loop in the middle of the loop body. The only thing goto should NOT be used for is complex flow control logic.

Variable names

Variable names are usually made up of lowercase letters; occasionally, digits and underscores are used when there's a good reason. Variable names should not be overly verbose.

Single-letter variable names are not frowned upon as long as their scope is reasonably short. For a function less than 20 lines, it's very reasonable for all local variables to have single-letter names. It's also reasonable for larger functions to use single-letter variable names for the main object they're working with for the whole duration of the function (e.g. "s" for a function that's processing a string).