Customer Reviews

UNIX to Linux® Porting: A Comprehensive Reference

5 star:		(3)
4 star:		(3)
3 star:		(0)
2 star:		(0)
1 star:		(0)

 

 

The cover announces that this is a "Comprehensive Reference". It certainly is, starting with a treatment of Linux development that almost could be a book by itself and ending with testing and development tools and techniques; it covers it all. In between are specifics of porting from Solaris, HP-UX and AIX, in great detail with plenty of advice and examples.

The only thing I'd question is the inclusion of so much unnecessary information. Is it necessary to have tables that show that "-o filename" is identical for both Linux and Solaris compilers (and of course two other tables show the same for HP-UX and AIX)? There is too much of that here; we could have saved a few small trees by leaving that out.

I suppose there are arguments to be made for including everything, but I'd ather concentrate on the differences. I can't fault the ators there: they give extensive coverage to the differences, and it's much more than just superficial listings
of flags or call arguments with differences (though again there's plenty of that too).

Overall, if I were porting from one of those OSes, I'd definitely want this book on my desk.

This book came out at just the right time to help with a large Solaris to Linux porting project. It's written by three IBM engineers who have a lot of experience doing Linux ports. There are a few introductory chapters that discuss many important things to consider in planning and scoping the porting project. These are followed by three chapters that go into great detail on what is involved in porting to Linux from Solaris, AIX and HP-UX (the three most popular UNIX variants). There is also a chapter that covers the use of Linux tools for application testing and debugging and a series of appendices. This is a very good collection of information to have all in one book. It would take quite a bit of digging through manuals and searching in the Internet to come up with information of this quality and detail. The detailed sections cover differences in compilers, linkers, debuggers, shell scripts, threading APIs, signals, system calls, libraries and other system facilities. I can't think of any area that the authors haven't addressed. The book definitely lives up to its subtitle, "A Comprehensive Guide." Note that it doesn't claim to be a "cookbook". Porting is too complex a task for that sort of approach. As and effective guide this book will save huge amounts of time and effort on the porting project and help make sure that you address all the important issues in planning the project.

One approach to porting that the book perhaps doesn't stress enough is a phased approach. The greatest advantage of Linux as a porting target is that all the development and debugging tools that are native to Linux are also available on UNIX. You can start by porting your applications to the GNU compilers, linker and assembler (binutils). All the necessary GNU development tools are available or can be built to run on a wide variety of UNIX platforms (and Microsoft Windows). You can get much of the porting work done without leaving your current operating system or hardware.

If you are not porting from Intel (or Alpha) based hardware, the endian byte ordering of your data could be a big issue. The book touches on this and gives some good pointers to places to look for help but I would like to see this issue addressed in more detail. It's a serious issue that can be the source of elusive runtime problems in a ported application.

Overall this is an excellent book.

This is a reference book, inasmuch as one is unlikely to read it cover to cover. It addresses a necessary but sometimes underappreciated point. That porting from a given Unix operating system to linux can be nontrivial. You probably know that linux tries to be like most unixes, at the command line level. Or indeed, at the structural level in how it partitions the user and kernel functionality. But just like porting from one Unix to another Unix involves a lot of gritty details, so too when decamping to linux.

In recognition of this, there are 3 long chapters, that describe going from Solaris, AIX and HPUX. Long on specifics. Like, consider pipes. A basic enabler of modularity at the command line. Every Unix has this. But did you know that Solaris pipes are full duplex, while linux pipes are half duplex? If you have a Solaris application, you'll have to pull apart some boundary code for its linux counterpart.

The entire book is like this! Nothing fundamentally difficult. Just a lot to check and possibly modify.

The book also has some early chapters which discuss the general aspects of porting from a Unix. It points out issues that you should consider, including a business purpose of why you want to port in the first place. These chapters should give you some clue as to the effort involved, if you decide to port.

This hands-on book deals with practical issues that arise when a C/C++/Fortran/Java application is ported from Unix to Linux. Specifically, the Unix variants the book deals with are Solaris, AIX, and HP-UX. Although the Unixes are generally the same operating system, they have their idiosynchrasies. I found this book very useful.

I highly recommend this book to anyone working in the Linux space. It's got everything you'll need.

The fact is, we shouldn't need this book. It is sad to see how many variants of UNIX are out there. This book covers converting programs that were developed and running on three of the major UNIX variants out there, Solaris, HP-UX and AIX, over to Linux. Such conversion entails not only fixing things that the compiler will complain about, like missing includes, incompatible system calls, etc., and also covers the more subtle incompatibilities, which might mean the program in question runs but crashes an hour later. It is an interesting reference work, and goes into great detail about trouble spots and such. I really hope that I won't see a similar book come out later, which covers converting programs written for several distinct variants of Linux so that they work on the next great OS. We should work together to make sure that such a balkanization does not occur.