Boost Testing :
From: Beman Dawes (bdawes_at_[hidden])
Date: 2005-10-20 20:34:25
"Jim Douglas" <jim_at_[hidden]> wrote in message
> My area of work is high reliability/high risk projects using the QNX
> RTOS and there is a "zero tolerance" approach to compiler warnings. If I
> derive classes from the serialization header files then the warnings
> will appear when I compile my application code and be rejected on QA
That's OK for compilers which allow supression of particular warnings via
command option, but it is hard for highly portable code like Boost's to
cater internally to every warning from every compiler. Effectively that
reduces the usable language way too much.
> I am new to Boost and I am willing to get more involved because I can
> see its potential applications. The question is, do the various authors
> look on their libraries as "industrial strength" utilities, or are they
> simply prototypes to test ideas for future inclusion into the C++
> Standard Library?
Very few Boost libraries are "simply prototypes to test ideas". Most are
intended for heavy-duty production use in a wide variety of environments,
and are in fact do get used that way a great deal. See the "Who is using
Boost" web pages.
Some concentrate more on providing a reference implementation than the most
all-out-efficient implementation, but we don't get a lot of complains about
speed, so I'm guessing efficiency is usually good enough.
But do understand that the implementation of each Boost library has its own
characteristics. So you need to evaluate each library for your particular
application, and avoid making assumptions about the rest of Boost based on
experience with a small number of the libraries.
PS: "Industrial strength" in my experience often means hurriedly developed,
underfunded, unreviewed, poorly tested, rarely documented, erratically
supported, etc. If that is your definition of "industrial strength" too,
then the Boost libraries will be a breath of fresh air.