The approach I've been using for a given header is to recursively do each of the "bits" headers which make up the standard header. So, e.g., while there are four headers making up , three of them were already documented in the course of doing other headers. "Untouched" means I've deliberately skipped it for various reasons, or haven't gotten to it yet. It /will/ be done (by somebody, eventually.) If you document an area and need to skip (for whatever reason) a non-trivial entity (i.e., one that should be documented), go ahead and add the comment markup, and use the homegrown @doctodo tag. See include/bits/stl_iterator.h for examples of this. Doing so will at least cause doxygen to consider the entitiy as documented and include it in the output. It will also add the entity to the generated TODO page. Area Still needs to be doxygen-documented ----------------------------------------------------------- c17 FINISHED (Nothing in Clause 17 "exists" in terms of code.) c18 FINISHED, Note A c19 Note A c20 Note A c21 Public functions basic_string done, Note B c22 Untouched; see docs/html/22_locale/* c23 See doxygroups.cc and Note B. Notes on what invalidates iterators need to be added. c24 stl_iterator.h (__normal_iterator, other small TODO bits) stream iterators c25 stl_algo.h (lots of stuff) c26 , , stl_numeric.h[26.4], Note A c27 ios_base callbacks and local storage basic_ios::copyfmt() std_streambuf.h's __copy_streambufs() " " _M_* protected memfns (data has been done) fstream and sstream protected members backward/* Not scanned by doxygen. Should it be? Doubtful. ext/* Some of the SGI algorithm/functional extensions. All of rope/hashing/slist need docs. __gnu_cxx Tricky. Right now ext/* are in this namespace. ----------------------------------------------------------- NOTES: A) So far I have not tried to document any of the headers. So entities such as atexit() are undocumented throughout the library. Since we usually do not have the C code (to which the doxygen comments would be attached), this would need to be done in entirely separate files, a la doxygroups.cc. B) Huge chunks of containers and strings are described in common "Tables" in the standard. These are pseudo-duplicated in tables.html. We can use doxygen hooks like @pre and @see to reference the tables. Then the individual classes do like the standard does, and only document members for which additional info is available. STYLE: stl_deque.h, stl_pair.h, and stl_algobase.h have good examples of what I've been using for class, namespace-scope, and function documentation, respectively. These should serve as starting points. /Please/ maintain the inter-word and inter-sentence spacing, as this might be generated and/or scanned in the future. vim:ts=4:et: