Re: IMPLICIT NONE (F2k8+/-)

Hello,

robin wrote:

<snip>

In old codes, you can be virtually certain that debugging
would be required because of the bugs that turn up with
IMPLICIT NONE.


Of course, a tool such as ftncheck, or the cross-reference
ability of many older compilers, reduces the value of implicit none.
With argument intents, modern compilers are sometimes able
to deduce use-before-store errors, which also reduces the value
of implicit none.

    Gerry Weinberg had something to say about bugs.
If one person finds x bugs in some code from a visual
inspection, and another person finds y bugs,
Weinberg had a formula (in terms of x and y) that predicts
the number of bugs actually in the code.


Does the formula take into account the fact that actual measurements
of older Fortran codes reveal that something like 90% of all bugs
were due to interface mismatches, and not the lack of implicit none?

That is, is there an accounting of the likelihood of categories
of bugs built into the formula?  Such language-specific help
would make the formula more useful.

I'm just asking.  Bayesian statistics are generally more able to make
predictions than simple frequentist statistics.  I can't tell, from your
description above, what sort of statistics the formula uses.

<snip the rest>

--
Cheers!

Dan Nagle
Purple Sage Computing Solutions, Inc.
.

Relevant Pages

  • Re: Zero Known Bugs
    ... Zero Known Bugs. ... end program output ... program test_kinds implicit none ... Surely you can cook up something that can choke every compiler out there? ...
    (comp.lang.fortran)
  • Re: IMPLICIT NONE (F2k8+/-)
    ... Gary L. Scott wrote in message... ... |If code needs to be debugged due to the addition of implicit none (as ... you can be virtually certain that debugging ... Gerry Weinberg had something to say about bugs. ...
    (comp.lang.fortran)
  • Re: fortran 90 compiler for linux
    ... > So all compilers have bugs. ... > In the case of commercial compilers ... ... > compilers to generate applications (all applications are ... for the cutups of their developers and the QA of their misartistry. ...
    (comp.lang.fortran)
  • Re: Whats stopping you from using Ada for your next commercial project?
    ... commonality between Ada implementations, not to find bugs per-se. ... the ACATS does not have a lot of tests that combine various ... So while the ACATS raises the quality of Ada compilers, ...
    (comp.lang.ada)
  • Re: fortran 90 compiler for linux
    ... So all compilers have bugs. ... In the case of commercial compilers like Lahey LF95, ... Standards related or not, are taken quite seriously, ... even less often related to Fortran 95 Standards. ...
    (comp.lang.fortran)