Checking code (was: Software engineer tech. writers? )

[Sandy Harris]

On 9/2/07, Dossy Shiobara <dossy at panoptic.com> wrote:

> Are there software-oriented tech. writers who are also software
> engineers who can read through C source and write documentation
> based on what it does?

How many writers analyse the code they document?

At one point I had access to 600,000 or so lines of a client's C code,
so I wrote a few little grep scripts to test it a bit. Results were appalling.
It was production code, but only used as internal test tools; it wasn't
what they shipped to customers.

Results I recall:

Under 15% of "switch" statements had a "default" case.

The "assert" macro was used about 5 times in several thousand lines
by one programmer, not at all by any of the others.

The C idiom for checking return value of functions that allocate resources,
fopen(), malloc(), fork(), etc. is something like

                if( (fp = fopen(...)) == NULL)

This was rarely used. In most cases, there was no test in following
lines either.

If you do check code this way, what should you do with your results?
I considered passing them on to the programming manager, but did
not do it. I figured as a contractor and "only" a tech writer, it was not
a good idea to stir things up that way.

-- 
Sandy Harris,
Nanjing, China