Re: How to learn software design
In article <Xns9CE6F037A68FFpaavo256@216.196.109.131>, Paavo Helde <myfirstname@osa.pri.ee> wrote:
James Kanze <james.kanze@gmail.com> wrote in news:02ae8929-f6c1-408f-8efd-
daf0cf97c4c1@l13g2000yqb.googlegroups.com:
On Dec 19, 10:00 am, ta...@mongo.net (tanix) wrote:
If you make your algorithm BASICALLY work, but think that some
things may NEVER happen, think again. Becasue they WILL.
Yah. Someone will turn the machine off.:-) (Seriously, if you
mean what I think you mean, then I agree 100%. Consider all
possible inputs. In a multithreaded environment, consider all
possible thread switches.)
I have hunted down and fixed a race condition error (one of my own,
unfortunately) in a multithreaded application which occured only after
running approximately 1 hour on a 16-core machine in parallel. I guess for
a 2-core machine it would have taken years to exhibit the bug.
Indeed, this episode taught me something new about "rare and unimportant
events".
Paavo
Well, I'll tell you even better than that.
One one of my contracts on the global voice messaging system
with as heavy traffic as you can imagine, I had a client/server
system. Client was written for windows, server for Unix.
I ran the load tests.
It all worked fine for DAYS, millions of messages exchanged,
and than boom. One funky message had one character wrong...
Try to fix that one, especially considering you have both parts
of network traffic, client end and a sever end.
After that, I am paying much more attention to those things
that are "impossible to happen".
--
Programmer's Goldmine collections:
http://preciseinfo.org
Tens of thousands of code examples and expert discussions on
C++, MFC, VC, ATL, STL, templates, Java, Python, Javascript,
organized by major topics of language, tools, methods, techniques.