OT Verifying code across gcc versions [Re: Commercial support on GCC/G++ open source versions]

[Paul Pluzhnikov]

Maurizio Loreti <mlo@foobar.it> writes:

> You have almost exectly described my work as a programmer in the last
> 36 years (I have started in 1969) -- apart from the fact that I don't
> have to depend from "other products, some of which may only be
> available in binary form".

So *all* code you depend on is available in source.
That must mean that you only use Linux, because on every other
platform you do depend on "other code available only in binary form"
if only for libc.

Also, this could not be have been true throughout your 36 year
experience, since Linux is only 15 year old.

> My work is feasible because, e.g. at FNAL, we use strict ANSI Standard
> C++.  

Huh? How does the fact that your code is strictly standard help you
verify that none of it is mis-compiled or mis-optimized?

> Nowadays nobody of us fears a new version of the compiler.

I wonder why? 

To be absolutely sure that a new version did not break anything,
you'd have to have a test suite with 100% condition, decision and
statement coverage, and have that test suite run cleanly with old
and new version. But such a test suite can not exist for any
moderately-complicated system, especially if said system is in
maintenance (i.e. is being modified daily).

So what gives you confidence to "jump to new version as soon as it
is released"?

Cheers,
-- 
In order to understand recursion you must first understand recursion.
Remove /-nsp/ for email.