> Do you realize what you are asking? You want us to review each line of
> code to see if it might be affected by this bug! Then you want us to
> replace each incident with something that will generate the division,
> instead of letting the compiler pre-calculate the constant and simply
> load it.
yes, it's definitely painful. but not necessarily impossible or
unreasonable, depending on various things such as the total size of the
project, the number of developers working on it (the higher the better),
etc. it sounds like you have a large product with few devs, though.
> We would not do such a thing. Instead we have reverted to a previous
> compiler. I have been critical of Metrowerks in the past and your
> response re-enforces my belief. It shows a lack of understanding
> of the development environment on the part of Metrowerks.
wow. because Metrowerks can't turn around in 24 hours and give you a fix,
they don't understand development?
switching to an older compiler is of course a perfectly valid response. but
the bug only affects division of constant floating point numbers. i would
have thought it would be very easy to come up with a regular expression you
can use to grep your entire codebase to quickly find all code affected by
this bug. yes, you might have #define's that obfuscate the grep. so just
grep for '/' and filter out all '/*' or '*/' or '//'. etc. doesn't seem
like rocket science to me, but probably there are other issues complicating
your situation.
> 1) In a mature design, not all lines of code are known by a single
> designer. To suggest that all code can be reviewed and simply fixed
> using a sub-optimal work around is preposterous.
i work on a very large commercial product for Windows with somewhere around
a million lines of code and 65 developers. we fairly routinely go through
our codebase making various workarounds or other kinds of maintenance.
recently we've begun the process of completely removing C++ exceptions from
our product (i don't want to get into the reasons why, just accept that we
have decided that for our purposes exceptions are a Bad Thing). in my
experience it's not that big of a deal to make sweeping mechanical changes
to a large codebase. in fact, it often helps new hires get a better grasp
of the product and its architecture. obviously, that's just my own
experience, and your mileage may vary.
