An eclass is a collection of code which can be used by more than one ebuild. At the time of writing, all eclasses live in the eclass/ directory in the tree. Roughly speaking, there are three kinds of eclass:

• Those which provide common functions which are used by many ebuilds (for example, eutils, versionator, cvs, bash-completion)
• Those which provide a basic build system for many similar packages (for example, vim-plugin, kde)
• Those which handle one or a small number of packages with complex build systems (for example, vim, toolchain, kernel-2)

Before committing a new eclass to the tree, it should be emailed to the gentoo-dev mailing list with a justification and a proposed implementation. Do not skip this step sometimes a better implementation or an alternative which does not require a new eclass will be suggested.

Before updating eutils or a similar widely used eclass, it is best to email the gentoo-dev list. It may be that your proposed change is broken in a way you had not anticipated, or that there is an existing function which performs the same purpose, or that your function may be better off in its own eclass. If you don't email gentoo-dev first, and end up breaking something, expect to be in a lot of trouble.

When removing a function or changing the API of an eclass, make sure that it doesn't break any ebuilds in the tree, and post a notice to gentoo-dev at least 30 days in advance, preferably with a patch included.

If there is an existing maintainer for an eclass (this is usually the case), you must talk to the maintainer before committing any changes.

It is not usually necessary to email the gentoo-dev list before making changes to a non-general eclass which you maintain. Use common sense here.

A simple way to verify syntax is bash -n foo.eclass.

No longer used eclasses may be removed from the tree, but developers must adhere to the following process:

1. Make sure that no packages or other eclasses in the tree inherit the eclass.
2. Send a lastrite message to the gentoo-dev and gentoo-dev-announce mailing-lists, announcing that the not-used eclass will be removed in 30 days.
3. Add a one line comment to the eclass, saying exactly # @DEAD so that the eclass-manpages package will not attempt to document it.
4. Wait for the 30-day period to pass, then remove the eclass from the tree.

An eclass is a bash script which is sourced within the ebuild environment. Files should be a simple text file with a .eclass extension. Allowed characters in the filename are lowercase letters, the digits 0-9, hyphens, underscores and dots. Eclasses are not intrinsically versioned.

Eclasses should start with the standard ebuild header, along with comments explaining the maintainer and purpose of the eclass, and any other relevant documentation.

Top level variables may be defined as for ebuilds. If any USE flags are used, IUSE must be set. The KEYWORDS variable must not be set in an eclass.

Eclasses may define functions. These functions will be visible to anything which inherits the eclass.

As an example, the following eclass was written to illustrate a better way of installing OSX application files during a discussion on this subject. It defines a single function, domacosapp.

An eclass may provide default implementations for any of the standard ebuild functions (src_unpack, pkg_postinst etc). This can be done either as a simple function definition (which is not multiple eclass friendly) or using EXPORT_FUNCTIONS. Functions given to EXPORT_FUNCTIONS are implemented as normal, but have their name prefixed with ${ECLASS}_.

If multiple eclasses export the same function, the latest (inherited last) defined version wins. If an ebuild defines a function that is exported, this gets priority over any eclass version. This can be used to override eclass-defined defaults for example, say we had fnord.eclass:

Then an ebuild could do this:

A simple eclass which defines a number of functions and a default src_compile for the (hypothetical) jmake build system might look something like the following:

Sometimes an eclass is used incorrectly by an ebuild and the eclass knows it is being used incorrectly- the common example is an eclass that only works with a specific set of EAPIs, but is being accessed inherited by an ebuild with a different EAPI. In those cases, used sparingly as a last resort, it is allowed for an eclass to invoke die from the global scope. For example: