wiki:Developer/Coding/Doxygen

Version 2 (modified by Sh, on 05/08/08 at 12:46:53) (diff)

Doxygen Recommendations

RTEMS is much older than Doxygen and the documentation in the .h and .inl files was obviously not written with Doxygen markup. In 2007, JoelSherrill undertook converting the documentation in the .h and .inl files in the RTEMS SuperCore? to Doxygen format. As a result of this effort, the Doxygen for the CVS version of the RTEMS SuperCore? is now built automatically multiple times per day and made available on the RTEMS Website. In April 2008, JoelSherrill began to update the Classic API (e.g. cpukit/rtems) .h and .inl files to include Doxygen markup.

Files

You can document a file with the @file [filename] directive. If you omit the filename Doxygen will use it for the documentation of the file containing this directive. This is the preferred way to document files which can be read by Doxygen. You should not use @file name_of_current_file to document the current file because this instructs Doxygen to document a file with the filename name_of_current_file and this is not necessarily the current one and may lead to ambiguity if someone from outside reuses your stuff to generate her documentation. Within one Doxygen run all files are unique and specified by the current Doxyfile. You can define how the generated output of path and filenames looks like in the Doxyfile via the FULL_PATH_NAMES, STRIP_FROM_PATH and STRIP_FROM_INC_PATH options.

Header Files

It is a RTEMS build feature that header files need to be installed in order to be useful. One workaround to generate documentation which allows automatic link generation is to use the installed header files as documentation input. Assume that we have the RTEMS sources in the rtems directory and the build of our BSP in build/powerpc-rtems4.9/mybsp relative to a common top level directory. Then you can configure Doxygen like:

INPUT           = rtems/c/src/lib/libbsp/powerpc/mybsp \
                  rtems/c/src/lib/libcpu/powerpc/mycpu \
                  rtems/make/custom/mybsp.cfg \
                  build/powerpc-rtems4.9/mybsp/lib/include/myincludes

RECURSIVE       = YES

EXCLUDE         = rtems/c/src/lib/libbsp/powerpc/mybsp/include \
                  rtems/c/src/lib/libcpu/powerpc/mycpu/include

FULL_PATH_NAMES = YES

STRIP_FROM_PATH = build/powerpc-rtems4.9/mybsp/lib/include \
                  rtems

Script and Assembly Files

Doxygen cannot cope with script (= files with #-like comments) or assembly files. But you can add filter programs for them (TODO: Add source code for filter programs somewhere):

FILTER_PATTERNS = *.S=c-comments-only \
                  *.s=c-comments-only \
                  *.cfg=script-comments-only \
                  *.am=script-comments-only \
                  *.ac=script-comments-only

Assembly Example

/**
 * @fn void mpc55xx_fmpll_reset_config()
 *
 * @brief Configure FMPLL after reset.
 *
 * Sets the system clock from 12 MHz in two steps up to 128 MHz.
 */
GLOBAL_FUNCTION mpc55xx_fmpll_reset_config
    /* Save link register */
    mflr r3

    LA r4, FMPLL_SYNCR

You have to put a declaration of this function somewhere in a header file.

Script Example

##
#
# @file
#
# @ingroup mpc55xx_config
#
# @brief Configure script of LibBSP for the MPC55xx evaluation boards.
#

AC_PREREQ(2.60)
AC_INIT([rtems-c-src-lib-libbsp-powerpc-mpc55xxevb],[_RTEMS_VERSION],[http://www.rtems.org/bugzilla])

Functions

For documentation of function arguments there are basically to ways: The first one uses @param:

/**
 * Copies bytes from a source memory area to a destination memory area,
 * where both areas may not overlap.
 * @param[out] dest The memory area to copy to.
 * @param[in]  src  The memory area to copy from.
 * @param[in]  n    The number of bytes to copy.
 */

This leads likely to copy&paste within the comment, bloats simple function comments and is not very fluent to read.

One alternative is this:

/**
 * Copies @a n bytes from a source memory area @a src to a destination memory
 * area @a dest, where both areas may not overlap.
 */

The @a indicates that the next word is a function argument and deserves some kind of high-lightning.

GCC Attributes

The Doxygen C/C++ parser cannot cope with the GCC attribute((something)) stuff. But you can discard such features with pre-defined preprocessor macros:

ENABLE_PREPROCESSING = YES
MACRO_EXPANSION      = YES
EXPAND_ONLY_PREDEF   = YES
PREDEFINED           = __attribute__(x)=

JAVADOC_AUTOBRIEF and QT_AUTOBRIEF

We should agree on either NO or YES for both.