coggins@retina.cs.unc.edu (Dr. James Coggins) (11/08/88)
Managing C++ Libraries: Table of Contents James Coggins and Greg Bollella Computer Science UNC-Chapel Hill Our sequence of articles on Managing C++ Libraries has reached five (six, counting the response to followups), and we're basically finished, so here's a Table of Contents of what we have posted so far. If you have not received all of the parts, e-mail to me (coggins@cs.unc.edu) and we'll be glad to ship them out by express e-mail. Problems in Library Management <4959@thorin.cs.unc.edu> ------------------------------ Provides some background on why we are producing these articles and introduces the problems we have encountered in managing our libraries. Provides brief introductions to the following 3 articles. Using +e0 and +e1 <4981@thorin.cs.unc.edu> ----------------- Shows how to use the Cfront 1.2.1 +e flags and describes their dramatic effect on the size of .a files and a.out files. Subdirectories and .c Files <5020@thorin.cs.unc.edu> --------------------------- Describes why and how we store our library in a directory hierarchy with each class in its own subdirectory. Actual Makefiles for the main library, a sublibrary, and a class are provided for use as templates that you can modify for your own library. We also describe the tradeoffs involved in storing each non-inline member function in a separate .c file (which we do and you might want to do). Dependencies and Headers <5078@thorin.cs.unc.edu> ------------------------ This is the most intricate article, but it may be the most useful. It shows how we manage the complex dependencies between classes and the related problems of proliferating header files. We describe approaches we have considered and rejected. We then show a scheme whereby dependencies can be declared and the right headers included with a minimum of intrusion on the application programmer or library developer. Minimizing Recompilations <5154@thorin.cs.unc.edu> ------------------------- We describe a simple scheme by which one can ensure that portions of the library that have undergone trivial changes will not be compiled when the next library or sublibrary compilation directive is issued. Responses to Followups <5078@thorin.cs.unc.edu> ---------------------- We respond in this article to some feedback we have received to this series. One class of comment involved "why not get a better preprocessor" and the other was "Here's another (simpler) scheme for organizing things." We don't buy the "simpler" part. The work described in these articles was supported in part by the National Science Foundation (through our department's SoftLab program), the Office of Naval Research, and the Universities Space Research Association: Center of Excellence in Space Data and Information Science (CESDIS).