32-bit Changes

John Tytgat John.Tytgat at aaug.net
Mon Apr 21 13:34:32 PDT 2003


In message <264ec9e64b.peter at chocky.org>
          Peter Naulls <peter at chocky.org> wrote:

> In message <cd74c6e64b.Jo at hobbes.bass-software.com>
>           John Tytgat <John.Tytgat at aaug.net> wrote:
> 
> > I'm always in favour having a set of sanity test cases checked in which
> > can be run to check if a certain fix did break something obvious.  Please
> > do !
> 
> I'm not proposing to do any more than I've done mind, but are there any
> other specific areas of concern you have?

Well, I'm wondering if there isn't any free set of a (Posix, ANSI, etc)
test set we can use to ensure a certain level of correctness of UnixLib
so that each time we do some changes in UnixLib, we can at least run the
same test set and ensure to have the same (if not better) output results.

If there isn't (or it is too big, or UnixLib isn't (yet ?) covering
enough for such a test set), we can perhaps collect our private test files
we have (I have a couple for testing the backtraces and signal code)
in a directory together with a small explanation what these files should
output.

> Incidentally, my autobuilder
> stuff serves as an excellent regression test for header correctness -
> 60 odd programs/libraries all using the same setup.  

That's ensuring to have this.  The problem is that when someone else
checks in a change, it is only you to pinpoint the problem.

John.
-- 
John Tytgat, in his comfy chair at home                                 BASS
John.Tytgat at aaug.net                             ARM powered, RISC OS driven



More information about the gcc mailing list