Could someone tell me why linux has many header files in the /usr/include direcory with an identically named file in the /usr/include/bits directory. For example, if one wants to see a definition of the termios struct it isn't in /usr/include/termios.h, but in /usr/include/bits/termios.h. Of course, there is an #include <bits/termios.h> statement in termios.h, but why are there two files?

htarko

Moved: This thread is more suitable in the Programming Forum and has been moved accordingly to help your thread/question get the exposure it deserves.

All standard C++ headers are supposed to be .h-less. The names of the headers and behaviors of the classes are specified in the C++ standard, however the implementation is left up to the lib author. Most of the headers are for templates and therefore include some source code, and many of the includes include multiple bits headers (such as constants and settings that have separate files.) Basically, the main directory is for the standard-compliant file names, and the bits subdirectory allows the developers to manage the code in whatever manner they choose.
ta0kira

The bits sub-directory is there to support diffent "endianness", for one thing.
In other words, it is system-specific. If you go to a big endian system or a little endian system the same filenames will be there. What's in them will be different.

The /usr/include/sys subdirectory does the same thing. You'll notice that stat.h is in sys because how it works is usually system-specific.

Thank you. Tha makes sense