A: First, remember that the preprocessor (cpp) #includes the iostream.h header file into your source file, before the compiler ever sees the file. iostream.h itself #includes other header files, etc., so several potentially large text files are added to your hello.cpp program before it's ever compiled.

We can use the g++ command to invoke cpp and save cpp's output to a file: g++ -E hello.cpp > pre-hello.cpp

Note the size of the pre-processed file, pre-hello.cpp: 24,894 bytes.

The file has been pre-processed but not yet compiled. Use the "more" command to examine the contents of the file to see what cpp added: more pre-hello

Next, use g++ to compile hello.cpp to the executable, hello: g++ -o hello hello.cpp

Note the size of the executable, hello: 705,576 bytes!

To understand what is going on, note that the compiler allocates space for an executable in three segments: Text, Data, and BSS.

The text segment stores executable statements, the data segment stores initialized variables, and the BSS segment stores variables that don't have a value yet. The size of each of these segments is given by the Unix size command.

Examples:

% size hello
98066 + 15532 + 1516 = 115114

% echo "text  data   bss   total" ; size a.out
text     data     bss    total
98098  + 15532  + 1516 = 115146

The rest of the answer can be determined by using the Unix strip command, which removes the compiler-generated symbol table and other debugging information from a.out. This information was added to a.out by the compiler, but it is not stored in any of the three segments-- it is extra information used only for debugging tools. Therefore, this command is normally run only on programs that have been debugged and tested.

% strip a.out

text     data    bss   total
98098 + 15532 + 1516 = 115146

Further Reading

• Books: For the complete story with pictures and diagrams, see pp. 138-149 in EXPERT C PROGRAMMING: DEEP C SECRETS, by van der Linden (Prentice Hall), a book that received a five-star review in Unix World. The UO library should have it.

• Online Manual Pages. a.out, g++, cc, strip, cpp, elf(3ELF)

• Header files. If you use the -H option with g++, it will echo the pathnames of header files that cpp #includes. Then you can use emacs, for example, to read a header file like iostream.h.