GCC with formal testing docs

[Ben Elliston]

On Wed, 2007-07-25 at 07:05 -0400, Robert Dewar wrote:

> > If you build the compiler with coverage instrumentation and run the
> > testsuite, you might get a shock.  It's not as well tested as you might
> > think.
> 
> If it gave anyone a shock to find out that the test suite did not
> provide 100% coverage, then that person is not very familiar with
> compiler technology. It is by no means SOP to try to get 100%

I guess I should have been more specific.  The point I was trying to
make is that the GCC testsuite was never intended to be a coverage
testsuite.  It is primarily a regression testsuite.

I have run the testsuite under an instrumented compiler and, if I
recall, the coverage averaged over all the files was about 40%.

> coverage testing of a compiler, and in practice for many reasons,
> very difficult (compilers often contain a lot of deactivated code
> that comes from defensive programming against errors, since
> compilers more than many programs routinely expect to be fed
> rubbish, and work hard to behave nicely when mistreated in
> this way :-)

If you're determined enough, it is possible to test all of that code,
too.

Ben