Re: GNATS survey

["Joseph S. Myers" <jsm28 at cam dot ac dot uk>]

On Mon, 5 May 2003, Wolfgang Bangerth wrote:

> Yesterday, I had meant to spend my day skydiving, but then we had 100% 
> cloud cover over central Texas and I ended up mailing with Giovanni about 
> the state of GNATS and a couple of suggestions he had. The upshot was: we 
> would like to go through all the C++/C/Optimization PRs and check them for 
> validity. Of public interest in this respect is that we also would like to 
> change one of the policies in assigning states to PRs. Here's what:

middle-end PRs probably also fit in much the same grouping - in terms or
general nature of the bug reports, and arbitrariness of which category the
PRs are in (and there aren't that many middle-end PRs).  And a fair number
of "c" PRs should be moved to "optimization" or "middle-end" if they don't
look anything to do with the C front end itself (just problems the user
found compiling C code).

> - thus, we would like to reserve the "analyzed" state for PRs that are
>   both confirmed and equipped with a small testcase
>   NOTE: we're presently talking only about C++/C/opt PRs.

Where confirmed includes not just that the compiler behaves as the user
claims, but that this is indeed incorrect: if the bug report is that code
is wrongly accepted/rejected, it is indeed wrong for the compiler to
accept/reject that code, as you note:

> - we will find a couple of PRs that we are not sure whether they are 
>   standard conforming or not; it would be great if some of the language
>   gurus could volunteer to look at some of them as we find them so that we 
>   could be sure that after the clean-up everything that is "analyzed" 
>   really is.

-- 
Joseph S. Myers
jsm28@cam.ac.uk