This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: GCC Release Status (2003-08-22)


>| (Given that GCC 3.4 will be a "we break nearly all of your (non-ISO) C++
>| programs" release, we may want to push it out the door relatively quickly
>
> I'm skeptical about that approach.  I believe that if GCC-3.4 is
> rock-solid enough and does good jobs, then people will embrass it.
> However, if it is released too quickly and we did not have enough time
> to fix outstanding issues, then people will just ignore it and we
> would miss audiance there. 
 
That is what happened with gcc 3.0. That was not a bad compiler, but people 
still didn't accept it because it broke their code. It took them quite a 
while until they realized that it is really _their_ code that's broken, not 
gcc, and that waiting for the next release is not going to improve their 
situation. I wouldn't be surprised if we had the same this time -- a good 
compiler without people using it.

On the other hand, my feeling is that presently
- the C++ part certainly has quite a number of things to fix; we get a pretty
  continuous influx of bug reports with regressions on mainline. I wouldn't
  want to go out the door in the present shape, but things have become much
  better since the new parser went in.
- there are presently about 50 C++ PRs targeted for 3.4. That is not
  incredibly many, and I'm pretty certain that most of them will be fixed in
  time.
- overall, I think the C++ part is not in such a bad shape. I started working
  on gnats about a year ago, and we had 570 open C++ reports back then.
  We're presently at 330. Subtract the 50 that will have to be fixed for 3.4,
  and we have a rather impressive improvement in this area. I think the C++
  maintainers should really be commended for this!
- I use mainline since May for my daily development. It's not that it would be
  totally broken.

In other words, if the C++ maintainers say they will have the time to fix the 
existing regressions, then I don't see why we should give _much_ more time to 
let things settle. 

W.

-------------------------------------------------------------------------
Wolfgang Bangerth              email:            bangerth@ices.utexas.edu
                               www: http://www.ices.utexas.edu/~bangerth/


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]