How to see labels in common block

[Andy Vaught]

On Tue, 15 Feb 2000, Jeffrey A Law wrote:

>  In message < 38A9B600.C069E005@moene.indiv.nluug.nl >you write:
>   > 
>   > >   As far as improving the testsuite, we might consider simply downloading
>   > > a million lines of code from netlib.  It isn't something that we can pack
>   > > up and distribute, but in those million lines, every silly thing that you
>   > > can do with fortran is done someplace.
>   > 
>   > No, I don't like that.  We should add testcases to the testsuite that
>   > really point to problems.  Otherwise it'll get much to expensive to run
>   > testsuites.  Furthermore - an observation I borrow from Mike Stump - a
>   > testcase is pointing at a weak spot in the compiler.  We might get it
>   > wrong again.
> I agree with Toon.
> 
> Our testsuite is a regression testsuite -- meaning it tests for bugs that
> have already been reported to make sure we don't mess them up again.  The
> tests are meant to be small, which keeps the compile/run time reasonable
> and makes the tests understandable, even to those not 100% familiar with
> the test itself.
> 
> We have a separate process for doing larger scale testing as part of our
> release process.  For example, we used lapack as part of our testing
> program for the gcc-2.95 release.
> 
> jeff

 
  While lapack is a large hunk of code, its written by guys who've been
writing fortran ever since it was invented and really know what they're
doing.  The style within that body of code is very consistent, which is
not so good for the sort of bulk testing that I was thinking of.

  As Tim pointed out, we'd probably end up arguing whether certain
constructs were legal in the first place.  Now, I think this funny and
correct at the same time, but I think most cases of spurious errors or
incorrect code could be resolved without too much difficult.

  One thing that my "million lines of code" doesn't have is the self-tests
that I believe lapack has-- I am thinking more along the lines of "the
compiler doesn't crash" as opposed to "the compiler generates correct
code". 

      Andy 

-----------------                        XOLD(K,IC,I)=
Andy Vaught               ....        DO ITERS=1, 10  XOLD(K,IC,I)
andy@maxwell.la.asu.edu   |  |   /CALLMSOLVE(A,B,X,I,ITERS,TOL)+(RANNYU(0)
Arizona State University  ======|WRITE(6,'(I5,2X,F12.6)')ITERS,TOL -HALF)
Tempe, Arizona USA        OOOOOO \ENDDORETURN PARAMETER(ZERO=1.D0)*TENTH*DELTA