Customer Reviews

The Science of Debugging

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

 

 

I admit that I bought this book fully expecting to be
"underwhelmed." Instead, I was pleasantly surprised as
to how much this work had to offer.

They begin the book with this interesting premise:
"One of the groups that is likely to discount this book is
the system architects camp, although they are the
ones that are most likely to be aided by understanding the
root causes of bugs and how they can be avoided." After
reading this (which appeared on page 6) I decided to mark

each page that contained an important architectural-level
observation or recommendation. By the end of the book,
I had nearly 40 such pages identified!

The work is well researched with good references and
footnotes. It is also rich with real-world examples of
horror stories and debugging nightmares. I found myself
filling the margins with recollections of my own, similar
experiences.

The book is not without faults. There seems to be a great
deal of repetition and there were several points where I
wished the book had been more carefully edited. I didn't
agree with everything presented and I think they actually
got a couple of things completely wrong. Finally,
there is at least one puzzle presented where the authors
challenge you to "find the bug" where even their answer
wasn't complete! [See page 307--an exercise for the reader.]

The author's aim was to discuss debugging but the book
contained as many Software Engineering observations as
general debugging tips. I believe that these general
observations and recommendations strengthen the work
considerably.

Any college course in Software Engineering would benefit
by including this book as required reading. Any project
lead organizing a new product development effort should
take an hour and skim though this book again (after
an initial, careful reading). Finally, any programmer
who wishes to build for debug-ability would do well to
review this work.

The book's one sentence synopsis comes directly from the
authors (page 109) "One of the most critical (and most
overlooked) aspects of debugging is to understand why
bugs occur and where they come from." I agree. This
book does a marvelous job to address that and much more.
I highly recommend it.

OK, this book isn't really bad. But it has very, very little to do with debugging. Most of the book is given over to general software engineering (requirements, design, coding and test). Perhaps 1/4 of the book is really about debugging, and that seems to consist of two things - logging and tracing - over and over again.

Another thing that I found that grated on my nerves was the incessant references to the holy grail (allegedly) of computing, the 'production server'. I also assume that this is implicitly a 'windoze production server', since the authors seem to be of the opinion that if you even think of installing some software or patch on the server, then you are doomed to have all sorts of problems. They don't even seem to consider the possibility of running tools over the network, without installing them on the sacrosanct server machines.

There is zero coverage of debug tools (they say "read the manual of your debugger"). They do include code coverage, memcheck, source browser and defect tracking as debug tools. I wouldn't.

If you want a book that DOES cover debugging (and also some general software engineering, but less overwhelmingly so), then take a look at "Why Programs Fail" by Andreas Zeller.

In reading the book, there were many great platform idependant ideas for debugging applications. I think it would have been more helpful if the book would have at least cursorily implemented some of those ideas in code (my preference being C/C++). Some examples of what I mean are it says that one of the oldest ways of debugging applications use a log. It even goes to say what elements to include to make it as useful as possible. I think it would have been better to include the code implementing a logger they describe.