readme

apue第二版每一章节的c语言源码

This is the README file by W. Richard Stevens (PART ONE) and Tom Hanlon (PART TWO).
PART ONE is not Linux specific. PART TWO is specific to compiling the "Advanced
Programming in the Unix Environment" library for Linux Redhat 5.0 (kernel 2.0.32).
PART ONE has not been changed from the original which was written in almost pre-
linux days. Nevertheless, in order to understand PART TWO, you must read PART ONE.


			PART ONE by W. Richard Stevens
author of "Advanced Programming in the Unix Environment"

Source Files
------------

	All the source files (including this file) assume tab stops every
4 positions, not every 8 (the default).  Just ":set tabstop=4" in vi,
for example.  Yo no habla emacs, but I'm sure its doable in emacs.

What to do
----------

0.	Read this entire file.  Read the "DISCLAIMER" file.

1.	If you're using either SunOS 4.1.x, SVR4, or BSD/386, copy one of the
	files "Make.def.sun", "Make.def.svr4", or "Make.def.44" into the file
	"Make.defines".  This file ("Make.defines") is included by all the
	Makefiles in the individual directories.  Edit "Make.defines" as
	appropriate for your system.  Note that *YOU MUST MODIFY* "Make.defines"
	with the pathname of the "libmisc.a" library, any special compiler
	flags, etc.

	If you aren't using one of these three systems, start with one of these
	three "Make.def.*" files and edit is as appropriate.  I am interested
	in receiving any new versions of this file for additional Unix systems,
	as long as I can include them in future releases of this tar distribution.

2.	If you're using either SunOS 4.1.x, SVR4, or BSD/386, go into one of
	the directories "lib.sun", "lib.svr4", or "lib.44" and type "make" or
	"gmake".  (Be sure to read the rest of this file first.)  This makes
	the library "libmisc.a" that almost every other program requires.
	This is where the error functions from Appendix B.2 reside, for example.

	Note that 90% of the files in these three lib directories are just
	hard links to each other, as the source file is identical between
	the three systems.

	If you aren't using one of these three systems, pick the directory that
	is closest to your system, and modify as required ...  The files with a
	link count of 1 (i.e., those that differ between the three different
	systems) will probably require the most work to port to other systems.
	Note also that the file "systype.sh" is used by the Makefile in these
	lib directories to try and determine which system we're on.  You almost
	certainly need to modify this shell script also.

3.	The file "Sourcefiles" lists which source file corresponds to each
	Program in the book.  Go into the desired directory and make whatever
	program you want.  There is not a master Makefile that make's everything.

4.	Each Makefile may contain entries for programs other than those that
	are distributed in this source release.  These are probably for small
	test programs that I wrote while writing the book.  Since these programs
	aren't in the book, and since it would take too much time to go through
	and figure out what each one does and document it adequately, they are
	not in this release.

5.	Please let me know of any generic problems you encounter.  I may be able
	to answer questions if you're having problems, but questions of the form
	"what does it take to compile X under FOOix" I probably can't answer,
	since I don't have access (and time) to try all these programs under
	every Unix system in existence.

Using make
----------

	Each directory with source code has a Makefile in it.  Each Makefile
includes "../Make.defines".  Unfortunately on all three systems the make
program is a little different:

	- Under SunOS I use "make", although GNU make also seems to work.

	- Under SVR4 I use "make".  I've never gotten GNU make to compile
	  correctly under SVR4.  Fortunately the vanilla SVR4 make is OK.

	- Under BSD/386 I use GNU make.  The standard BSD make is quite different
	  and would require that the include line at the beginning of every
	  Makefile be different.  (Guy Harris posted patches to the BSD make
	  that let it handle "normal" include lines, to one of the Usenet
	  newsgroups in May 1992.)

POSIX.1 Compatibility
---------------------

	Be aware that *none* of the systems used (SVR4, 386BSD, SunOS 4.1.2)
is POSIX.1-1990 compatible--they all conform to the earlier 1988 POSIX.1
spec.  This means that you will get some compiler warnings for some of
the files.  The main difference between the two is that the 1990 POSIX.1
standard introduced the ssize_t return value for read and write.  The
SVR4 function prototypes in <unistd.h> still say "int write(...)".

	*YOU WILL ALMOST CERTAINLY* get a compiler error on the "ourhdr.h"
file the first time you try and compile anything, unless your system
supports the ssize_t primitive system data type that appeared with the
1990 POSIX.1 standard.  I've just added the typdef of ssize_t to the
appropriate headers on my systems.  Without root privilege you need
another solution.  The easiest solution might be to typedef ssize_t
in "ourhdr.h" to be an int.

C Compilers
-----------

	The only ANSI C compilers that I've compiled these programs under
are GNU C (Versions 1.39 and 1.40) under SunOS and BSD/386, and the SVR4
Standard C Development Environment 5.0.  I have not tried Sun's ANSI C
compiler (acc) on the Sparc.



			PART TWO - Linux Redhat 5.0 specific (kernel 2.0.32)
				by Tom Hanlon - [e-mail]


What is New
-----------

	The code you have unpacked is exactly the same as originally distributed
in 1992 except:

	a) The file /apue/Make.def.rhlin
	b) This /apue/README which includes all of the original README plus this 
	   PART TWO.
	and
	b) The directory /apue/lib.rhlin and its contents which have been edited
	   to adapt the library to linux.

What you need to do
-------------------

	a) Read PART ONE.
	b) Edit /apue/Make.def.rhlin so it contains the proper directories to
	   store your library in after it is compiled.
	c) Copy Make.def.rhlin to the new file Make.defines (that is in the apue
	   directory at the shell do %cp Make.def.rhlin Make.defines).
	d) Switch to the apue/lib.rhlin directory.
	e) At the shell execute the command "make" (that is %make). I usually
	   do a "make > op 2> eop" so that standard output is stored in the
	   new file op and standard error is in new file eop. That way if
	   you have problems you have a file to print as documentation to 
	   show to your expert.
	f) That is it if you are using rh5.0 and kernel 2.0.32 on an intel
	   machine.
	NOTE: this compiles on my very generic Pentium 166 with one warning and 
	      no errors. That is it compiles all of the *.c files in the
	      /apue/lib.rhlin used in the Makefile into the libmisc.a 
	      library for your use in compiling and using the programs
	      in Mr. Stevens' book.
	g) You are then on your own to start at the beginning of the book and 
	   compile and use each program as you go. If you have read PART ONE
	   you now realize that there are source code references in the individual
	   Makefiles that are not in the book. You will need to edit the Makefiles
	   as you go.
	NOTE: this library compilation is certainly not guaranteed to compile
	      all programs in the book perfectly. Indeed, it is guaranteed not to.
	      I am learning as I go, too. But, if you have gotten this far,
	      than you are just as well equipped to become your own expert as am I.
	      Feedback is greatly appreciated.   

General Comment and Opinion
---------------------------

	Just so you do not think this library will not work some what reasonably - 
consider the following - I did compile all of the programs in the /file and /mycat
directories with just minor editing to the Makefile and they all compiled without 
warning or error. And I tested some of them which ran exactly as described by 
W. Richard Stevens.
	If you are delving around in the source code in the apue/lib.rhlin c files 
please note that I tried to remember to put a /* th */ (my initials) in 
every place where I edited something.
	Final comment is based on my notes about linux from two books: "Linux
Applications Development" Johnson and Troan (1998) and "Linux Kernel Internels" 
Beck et al (1996). If in doubt, read the book. These are only my notes. 
	Linux follows POSIX standards but does implement classic System V IPC 
semiphores, message queues, and shared memory (not covered by POSIX) without 
STREAMS but uses stackable line disciplines and network kernel drivers (the
last seven words are Johnson and Troan's not mine) in lieu of STREAMS for speed.
Linux generally follows System V practice where there is no POSIX standard yet
with two exceptions. The first is BSD sockets which linux follows ( we are talking
about networking here). And linux also does BSD ptys. This paragraph is 
practically useless. It deserves about a dozen books of its own. 
	

END OF README